This PDF file contains the front matter associated with SPIE Proceedings Volume 12724, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

In this paper, the methods and technologies such as computation model, spatio-temporal big data analysis, propagation dynamics system modelling, scenario simulation calculation experiment, spatio-temporal statistical measurement, unbiased estimation of fault samples, hazard assessment of THE LATEST EPIDEMIC PANDEMIC, the spread of THE LATEST EPIDEMIC PANDEMIC, analysis and judgment of different epidemic risks and situations across the country, as well as the analysis of THE LATEST EPIDEMIC PANDEMIC situation abroad and the tracking study of epidemic situation in key countries are comprehensively used for analysis. The paper showed that making full use of computation model, mathematical models, big data, artificial intelligence and other information-based means to control the complexity of the spread of the epidemic situation, so as to reduce the high uncertainty of the analysis, research and judgment of the epidemic situation, is a major scientific issue for rapid emergency decision-making under the crisis of major outbreaks of disease and requires interdisciplinary scientific research.

Attention problems are common problems in adolescents and even adults. In this study, a training protocol based on flashcard game paradigm was designed, and reaction time and eye movement data were recorded during the training. The subjects received attention training twice a week for one month. Subjects took the IVA-CPT attention test before and after training. The study compared and analysed parameters such as reaction time, fixation distribution, fixation time, scanning length, etc. The experimental results showed that the task reaction time of the subjects was significantly shortened after attention training, and various parameters of eye movements were improved and statistically different. After training the attention quotients of IVA-CPT test were greatly improved and statistically different from those before training. This study provides a theoretical basis and technical support for the study of attention training and training effects.

Postoperative infection is clinically considered to be the leading cause of patient death within one year after kidney transplantation. Pulmonary infection accounts for a large proportion. The factors that cause lung infection after kidney transplantation are so complex that clinicians have difficulty in determining whether a patient will be infected. Existing methods are to statically predict whether the patient will be infected by using the various indicators of a patient before surgery. Aiming to efficiently predict the infection risk of patients at any time, this paper proposes a joint prediction method based on recurrent neural network. We collect 656 real clinical data of kidney transplantation patients, including the relevant index data of patients before kidney transplant operation and the index data of regular testing after operation. Firstly, it pre-processes and cleans clinical patient data. Then, it adopts a multi-layer perceptron (MLP) to extract the static features of preoperative data, and a GRU network to extract the features of postoperative examination at each time sequence node. Finally, a residual connection is used to fuse these two kinds of features extracted respectively from the preoperative and postoperative examination data. The fused features, containing both semantic information and temporal information, are fed into a neural classifier to predict the pulmonary infection risks of patients. The experiments with performance comparisons verify that our method is effective. Hope our method can help clinicians to predict postoperative infection in advance.

Anti-PD-1 have demonstrated significant clinical efficacy in tumor by reversing T-cell dysfunction and exhaustion, therapy enhancing anti-tumoral properties. However, the overall response rate of anti-PD-1 is not high, and responder experience tumor relapse within 2 years. The combination of targeting sphingosine kinase-1 and programmed cell death 1 as a potential therapy way to overcome the problems of immune checkpoint resistance and non-response. Therefore, the primary goal of this study was to discovery natural compounds for dual targeting sphingosine kinase-1 and programmed cell death 1 ligand 1. A natural product database was constructed, and ADMET was used to evaluate natural products for primary screening. Molecular docking and Molecular Mechanics/Generalized Born Surface Area found candidate natural products, three natural products docking conformation were performed 100 ns molecular dynamics simulation. Natural compound 2, 3, 4 have further study value as dual targeting Sphk1 and PD-1 inhibitors.

Amplitude-integrated electroencephalography (aEEG) is widely adopted for recognizing neonatal neurological disorders in clinics. Previous work has mainly analyzed aEEGs from a time series perspective, while clinicians are more concerned about images. This paper studies the efficacy of image representations in aEEG interpretation. To this end, we employ Amplitude-frequency contour map (AFCM) to express information on local amplitude distributions and overall temporal trend. During its generation, Multi-scale Box-Cox transformation is introduced to flexibly control the degree of amplitude compression and capture details in different amplitude regions. Moreover, one- and two-dimensional features are extracted and combined to enhance model predictions. The experimental results show the potential of image representation for aEEG recognition. This method can provide new ideas to analyze time series signals like aEEGs from the perspective of combining time series and image.

With the progress of society and the continuous improvement of science and technology, people pay more and more attention to disease prevention. Under the background of the vigorous development of medical and health undertakings and the intensification of population aging, the nursing work of clinical departments in hospitals is facing great challenges. This article has carried out targeted analysis and discussion with orthopedic clinical nursing service as the research object. According to different populations, it has adopted differentiated strategies to help them establish healthy lifestyle and nursing concepts. This article also combined with clinical needs to carry out relevant work for patients to improve nursing satisfaction. At the same time, on the basis of artificial intelligence, this paper designed the optimized configuration of nursing services, and tested the configuration function. The test results show that the convergence rate of orthopedic clinical nursing service framework is more than 70%. This shows that the framework can efficiently help users complete clinical services, improve the quality of clinical services, promote the healthy development of patients' rehabilitation and reduce the occurrence of doctor-patient disputes, so as to promote the better and faster development of medical and health undertakings.

The research data were obtained from the information of ER antagonists provided by the D problem of Huawei Cup. We screened the molecular descriptor variables to construct quantitative prediction models for the biological activity of the compounds against ER, and the classification prediction models for Caco-2, CYP3A4, hERG, HOB, and MN of the compounds, respectively. We also determined the optimal range of molecular descriptors that could lead to compounds with better bioactivity for ER inhibition and better ADMET properties. In turn, we provide data analysis and predictive models for breast cancer research. In the first step, we used the feature es election method to remove redundant variables, and applied random forest to rank the variable sinterms of relevance to lterout7 molecular descriptor variables: MDEC-23, max HsOH, minHBa, min HsOH, minHBint4, C1SP2 and nHBAcc. In the second step, the pIC50 values in the training set were used as dependent variables and the seven molecular descriptor variables were used as independent variables, and extremely randomized trees were applied to construct the non-linear regression quantitative prediction model between compounds and ER bioactivity. In the third step, we used different integration methods to construct classification prediction models for Caco-2, CYP3A4, hERG, HOB, and MN, respectively. In the fourth step, after eliminating the 1974 compounds with the sum of ADMET indicators less than 3, we inverse solve the prediction model using the particle swarm optimization algorithm to obtain the maximum value of pIC50 and the optimal solution for each molecular descriptor variable. pIC50 has a maximum value of 8.9055, and the molecular descriptors C1SP, minHBa, min HBint4, minHsOH, maxHsOH, nHBAcc, MDEC, nHBAcc, and MDEC-23 are 0, - 7.75978169, -3.42842314, 9.46164889, 0.51061438, 2.63674280, and 80, respectively.

The extraction of features from the electroencephalography (EEG) of patients with major depressive disorder (MDD) and the subsequent construction of classifiers using machine learning methods to fuse decision information from multiple classifiers at the decision level is an effective method to improve the final classification performance. However, the current multi-classifier fusion approach is plagued by the output error probability of the support vector machine (SVM) affecting the accuracy of the final decision. Thus, to counteract the influence of the error probability of the classifier outputs on the classification results, this study proposed a decision-level strategy based on SVM classification confidence. Specifically, multiple classifiers were trained and the confidence of the classifier classification outputs was constructed as confidence feature vectors to be input into the SVM for the final decision. Compared with a single classifier or the traditional multi-classifier fusion method, the proposed method exhibited a stronger ability to identify depression with an accuracy of 86.79%.

Stroke patients commonly suffered from motor impairments, and gait analysis is crucial for the effective guidance of poststroke rehabilitation. However, most existing 3D gait analysis systems require the attachment of optical markers to the body, and the systems are costly and spatially intensive. These restricted the use of the systems in general hospitals and communities. To address this issue, this study developed an affordable and markerless system to examine the gait characteristics of stroke patients, which was constructed by depth camera of Microsoft Azure Kinect, instrumented walkway, and analysis software. Spatiotemporal and kinematic parameters were measured directly, while kinetic parameters were derived via inverse dynamics using a 14-segment human body model. Sixteen healthy individuals and eight stroke patients were recruited for tests. The results showed that due to weakened lower limb strength in stroke patients, their peak joint moments, joint angles, and joint angular velocities decreased compared with healthy individuals. Consequently, stroke patients walked with reduced stride length, cadence, and single support time. Furthermore, the differences between the paretic and non-paretic sides of stroke patients were primarily observed at the knee joint (peak joint angle, angular velocity, and moment were significantly smaller compared to the non-paretic side) during swing phase. This suggests that rehabilitation training for stroke patients should place importance on knee joint. These findings demonstrate the significant potential of the present system in the gait analysis for stroke patients.

PLLA microspheres have a wide range of applications in the biomedical industry, for which there is a large demand for size uniform PLLA microspheres. While microfluidic technology has the advantage that the prepared microdroplets are uniform in size and process is controllable. In this paper, we aimed at microfluidic technology, and then carried out structure improvement design and Simulation of PLLA microdroplet fabrication for flow-focusing droplet generator aided by COMSOL simulation software. The simulation results show that the size of microdroplets decreases with the reduction of the width of the indented section, and the generation frequency of microdroplets also increases. While the variation of the length of the indented segment had little effect on the generation of microdroplets. On the basis of the simulation results, PLLA microspheres were fabricated in a practical experiment using the data obtained from the simulation. The results showed that the size of the microdroplets prepared by the microfluidic method was 60 um, with a CV value of 2.2%, and the size of the cured PLLA microspheres was 23 um, with a CV value of 2.5%. Compared with the conventional method, the microspheres prepared by this method have good size uniformity and the process is controllable. Thereby, the application of microfluidic technology in the biomedical industry is advanced.

Diabetes is a metabolic disease characterized by high blood sugar. Diabetes is prone to various complications that can seriously affect human health and even endanger lives. Currently, diabetes is tested in an invasive manner, which is prone to patient resistance and even leads to trauma infection. In response to the need for non-invasive testing, this thesis proposes a non-invasive sweat glucose detection system, which consists of a glucose colorimetric sensor, an industrial camera-based image acquisition system and a color recognition algorithm based on a MATLAB program. The effect of the sample volume of glucose solution on the experimental results was explored, and a linear detection range of 0-4.4 mg/dL was obtained.

As a kind of natural plant medicine, polygonum aviculare has diuretic pharmacological effect, but its mechanism is not clear. In this study, the network pharmacological method was used to study the diuretic pharmacological mechanism of polygonum aviculare. The main research methods are as follows: the ingredients of polygonum aviculare mainly came from TCMSP database and one paper; The targets of chemical composition were from HIT database; The targets of disease were from Genecards database; The protein-protein interaction data was from STRING database; software Metascape was used for enrichment analysis of targets; The bio-network association maps were drawn using Cytoscape software. In this study, total of 33 oliguria effective targets were obtained. The network diagrams between ingredients and targets, PPI action network diagrams and network diagrams between ingredients, targets and pathways were built. This study suggested that the core targets of PA anti-oliguria were TNF, IL6, VEGFA, MMP9, CRP and IL2. After consulting the literatures, this study found that the following 3 ingredients have diuretic activity in total, which are myricitrin, luteolin and gallic acid.

Gastric cancer (GC) is one of the most common cancers in the world. Although the incidence and mortality rates of GC have declined globally in the past decades, its prognosis is still poor. Moreover, the exact mechanism of GC has not been thoroughly studied. This study aimed to identify central genes to improve the prognostic prediction of GC and construct a regulatory network of messenger RNA (mRNA), microRNA (miRNA), and long noncoding RNA (lncRNA). Six genes (COL10A1, CTHRC1, FAP, FNDC1, INHBA, and SULF1) were found using the information in the GEO database. The expression of these genes differs between patients with GC and nontumor groups, and such difference may affect the survival rate of patients with GC. A ceRNA network consisting of mRNA (CTHRC1, FNDC1, and INHBA), miRNA (mir-9-5p), and lncRNA (LINC00665) was constructed to reverse predict the target genes related to GC prognosis. According to the ceRNA theory, LINC00665 could bind with mir-9-5p in a competitive manner, increasing the transcription of CTHRC1, FNDC1, and INHBA regulated by mir-9-5p. However, this study has many limitations. First, age, sex, tumor stage, patient classification, and other factors were not considered when examining differentially expressed genes. Second, only FNDC1 expression in mRNA was shown.

The lack of medical resources per capita in China has led to many difficulties in the early prevention and control of diseases. In order to improve the effectiveness of early diagnosis of adverse pregnancy outcomes, improve the efficiency of medical services, this paper designs and implements a clinical decision support system for adverse pregnancy outcomes based on distributed computing and artificial intelligence technology. Visual workflow orchestration simplifies the process of collecting personalized clinical data by healthcare professionals; Distributed computing greatly improves the efficiency of clinical data extraction; The use of artificial intelligence models for early risk assessment and causal analysis of adverse pregnancy outcomes has achieved better results.

A method for the determination of tiopronin content in drug by copper(І)-cuproine spectrophotometry is established. The hydrosulfuryl(-SH) in tiopronin can quantitatively can reduce Cu2+ to Cu+, Cu+ react with cuproine to form copper(І)- cuproine amaranth complex in the condition of pH=5.0, the content of tiopronin can be determinated by measuring the absorbance of copper(І)-cuproine amaranth complex. The optimum conditions for the determination of tiopronin by copper(І)-cuproine spectrophotometry are discussed. Under the optimal conditions, the content of tiopronin in drug is determined, and the results are satisfactory.

This paper proposes an improved digital microfluidics (DMF) chip based on a printed circuit board (PCB) to realize the precise control of droplets. Solder mask layer was introduced to the chip as the dielectric layer, and this way can simplify the preparation processes of the chip, and the chip offers reasonable control of the droplet. Combined with the external module, the S protein peptide was screened on the chip. As a result, this PCB-based chip provided a new tool for droplet manipulation. It may have potential applications in drug screening, especially in the field with many repetitive operations.

This paper presents a novel algorithm for measuring tidal volume on a ventilator at different altitudes. The algorithm is based on real-time pressure detection and compensates for altitude variations to accurately calculate tidal volume and ensure the precise delivery of air. An experiment was conducted to compare the measurement error and consistency of this method with conventional methods. The results showed that using an elevation-based tidal volume measurement algorithm resulted in less error and better consistency of measurement results at different altitudes. The proposed method is precise, reliable, and has the potential to improve the accuracy of ventilator air delivery, reducing the risk of harm to patients.

With the progress and development of science and technology, robotics technology is increasingly used in the intervention treatment of children with autism spectrum disorders to help them grow better. This article discusses the application of robotics in the treatment of autism by combing related research at home and abroad, and looks forward to future research.

As the incidence of type 2 diabetes increases and tends to be younger, the probability of complications caused by it is also increasing, and people's healthy and safe lives are seriously affected. In this paper, based on 200 cases of data from the National Population Health Science Data Center's "Shared Cup Version_Diabetes Complication Early Warning Data set", we conducted preliminary screening, feature extraction, and selected three data indicator variables of blood pressure, blood glucose, and blood lipids for the study; constructed LSTM models, trained and tested the processed data, and then evaluated and tested the model prediction effect using confusion matrix and ROC curve test. After the model evaluation and testing, the prediction results were as follows: blood pressure: accuracy: 82.50%, precision: 92.00%, recall: 82.14% and F1: 86.79%; blood glucose: accuracy: 77.50%, precision: 86.21%, recall: 83.33% and F1: 84.75%; blood lipid: accuracy: 82.50%, precision: 93.75%, recall: 85.71% and F1: 89.55%. The model prediction effect of this experiment is obvious. The analysis and prediction of medical type 2 diabetes combined with coronary heart disease by establishing LSTM model provides a practical basis for early diagnosis and treatment of type 2 diabetes combined with coronary heart disease, thus better reducing the disease risk of cardiac function.

Activation of C-H is of great significance in the construction of complex molecular scaffolds in biomedical field. In recent years, enzyme catalysis has received widespread attention due to the high regio- and stereoselectivity and conforms to the current low-carbon policy, more and more reactions realize the non-natural C-H activation by reforming P450s. However, addition of NADPH cofactor will increase costs and may cause side effects. In order to circumvent the dependency on the reductase, we utilized biocompatible g-C₃N₄ porous nanosheets modified with P450 BM3 to form a simple, green and low-cost enzyme-photocoupled biocatalytic system for efficient solar-driven C–H oxidation. Finally, 25 μmol/L 11- hydroxy laurate was produced by photocatalysis of lauric acid, P450 BM3@C₃N₄ and TEOA for 3 h.

Pancreatitis is a serious pancreatic disease. Due to its inconspicuous early symptoms, it is easy to be ignored or misdiagnosed, causing many patients to miss the best time for treatment, and even life-threatening. According to statistics, the prevalence of pancreatitis is increasing year by year, and it has become one of the diseases that endanger human health. Therefore, early warning has become an important link in the treatment of pancreatitis, and laboratory test data is a reliable predictor. In this study, an early warning system for laboratory data was constructed based on neural networks to predict the occurrence of pancreatitis. The neural network model uses a 3-layer feed-forward neural network and uses the backpropagation algorithm for training. The experimental data comes from the laboratory test results of 90 patients with pancreatitis. Predictive indicators included 24 blood cell analysis data. Through the model verification experiment, this study achieved good prediction results, proved the feasibility and prediction accuracy of the model, and provided ideas and basis for the practice of computer-aided diagnosis of late pancreatitis. In view of the limited sample data, the experiment only The two-category research is limited. In order to better expand, the sample data will be expanded later to cover common diagnostic types and improve the generalization ability of the model.

Objectives: By the ways of Network Pharmacology and Molecular Docking Technology，researching the potential targets and the mechanism of action of proanthocyanidins combined with Allicin against atherosclerosis(AS). Methods: Searching in the Genecards database, collect the targets of Proanthocyanidins, Allicin and AS and try to analyse the common targets of them. Import the common targets of Proanthocyanidins, Allicin and AS into the web platform called Venny2.1.0 and get the Venn diagram; then after importing the intersecting targets that we got into the STRING database to build up the PPI net, beautify the PPT net by the Cytoscape software which could lead to the target nets of “Medicine and Disease.” Import the targets of how the medicines is used to treat the AS which is analyzed before into the web platform Metascape to perform GO and KEGG enrichment of analysismedicines and diseases. Make use of vina, pymol, ligplot and GROMACS software to perform the Docking validation and Molecular dynamics studies about the affinities among the medicine ingredient targets and core AS targets. Results: The key targets of Proanthocyanidins combined with Allicin against the relevant proteins of AS include Jun proto-oncogene(JUN), epidermal growth factor receptor(EGFR), mitogen-activated protein kinase 3(MAPK3), tumor necrosis factor (TNF), interleukin 6(IL6) and vascular endothelial growth factor A (VEGFA), etc. According to the enrichment analyses of GO, we find 2 pathways: GO:0000302 response to reactive oxygen species and GO:0009725：response to hormone; According to the enrichment analyses of KEGG, we find that AS closely relates to cancer, Lipids and Fluid shear stress. Docking validation and Molecular dynamics studies show that Proanthocyanidins combined with Allicin are capable to stably combine with eNOS and TNF-α. Conclusion: Proanthocyanidins combined with Allicin can fight against AS by IL-6, TNF, VEGF and JUN and other core targets. Meanwhile, by the pathways of Cancers, Lipids and Fluid shear stress.

For the current machine learning prediction model, there are problems such as low prediction accuracy of the pharmacokinetic properties of breast cancer candidate drugs, too many parameters of the neural network prediction model, and long training time. In this paper, we propose a lightweight multi-layer perceptual ensemble learning model based on imbalanced datasets to classify the pharmacokinetic properties of drug candidates. Experiments have proved that the proposed model is about 7% more accurate than the traditional machine learning model and 3% more accurate than the current advanced mainstream neural network model under the same public data set.

Prodigiosin is a kind of tripyrrole cyclic pigment produced by diverse microorganisms with a wide spectrum of biological activities and extensive applications. It is a promising antibacterial drug. Our research aimed to identify the antibacterial effects of prodigiosin against S. aureus. The minimum inhibitory concentration, changes in nucleic acid content and protein content, as well as the microstructure of cells were detected to reflect the results of the action of prodigiosin. At the same time, data analysis and visualization of related literature on bacteriostasis of prodigiosin were carried out by bioinformatics method. The results indicated that the MIC of prodigiosin on S. aureus was 0.075 μg/mL, and the cell membrane permeability of S. aureus increased after the action of prodigiosin, and the morphology of the bacterium was significantly wrinkled and broken. It is concluded that prodigiosin has the potency as an antibacterial agent against S. aureus. It provides an experimental basis for prodigiosin to become a new antibacterial drug. The bibliometric analysis in this study would provide a reference for researchers, and support the development of prodigiosin in the future.

With the continuous development of big data and artificial intelligence in the medical field, Radiomic which is a new imaging technology plays an increasingly important role in the diagnosis and prognosis evaluation of diseases and the analysis and prediction of tumor gene expression. This study provides a bibliometric and visualized analysis of the 100 most frequently cited articles in the field. A total of 9,780 articles were retrieved, of which 6,402 were of the Article type, among which we selected the 100 most frequently cited. The method reveals the development trend of this field to provide an effective reference for the development of Radiomic.

Rectal cancer poses a huge threat to human health. To establish appropriate treatment strategies, precise preoperative staging is essential. Computed tomography (CT) has consistently played a crucial role in the preoperative examination of rectal cancer patients. However, the diagnostic abilities and efficiencies of radiologists in identifying rectal cancer T stages using CT images still need improvement. With the rapid advancement of deep learning (DL) technology, applying DL technology to CT image recognition can expedite the construction of an efficient image recognition platform, offering novel possibilities to tackle this issue. In this study, a novel ResFusNet model is proposed for rectal cancer T staging using CT images, showing efficient and accurate performance. For the ResFusNet model, the accuracy, precision, recall, F1 score and Matthews correlation coefficient (MCC) of the test set reached 99.89%, 99.94%, 99.84%, 99.88% and 99.85%, respectively. These results notably outperformed the performance of other models. In light of this, ResFusNet has the potential to become a highly sensitive model for rectal cancer T stage diagnosis and serve as a valuable auxiliary tool for clinical physicians

A finite element model was created to analyze the mechanical response of the lumbar spine, dura mater, and cerebrospinal fluid (CSF) under fluid-solid coupling. This model provides a theoretical basis for studying the impact of CSF flow on the spine. The method utilizes CT scans of patients with Lenke3 scoliosis and employs medical image processing and 3D modeling software to create a solid-fluid coupling model of the lumbar spine's CSF. The simulation operations involve finite element software, which includes meshing, material property determination, contact definition, and boundary condition setting. The simulation results showed the distribution of stress in the dura mater and 5 lumbar vertebrae. The outlet position for CSF flow typically corresponded to the location of the highest velocity vector. At flow rates of 5, 8, and 12 cm/s, the corresponding results were 16.23, 25.32, and 47.2 cm/s. The maximum Von Mises stress values for the dura mater at three flow rates were 1.634x10^-3 Mpa, 3.699x10^-3 Mpa, and 7.456x10^-3 Mpa. The force of CSF flow on the dura mater is transmitted to the lumbar vertebrae, resulting in varying Von Mises stress magnitudes and distributions across each segment. The L2 segment experiences the highest stress levels at all three flow rates, with maximum Von Mises stress values of 1.598x10^-4 Mpa, 3.596x10^-4 Mpa, and 7.18x10^-4 Mpa on the lumbar vertebrae. As the CSF flow rate increases, so do the maximum velocity vector value at the outlet position and the Von Mises stress values on the dura mater and lumbar spine. The CSF flow rate can impact lumbar bone cell proliferation.

The purpose of the present study is to evaluate the biomechanical differences between two implant-supported cement retained single crown, screw-retained single crown and screw-retained union crown in premolar area. Thus, three prosthetic restoration groups were established based on different abutment types and crown retention designs: 2 straight abutments and 2 single cement-retained crowns (T1), 2 screw-retained abutments and 2 single screw-retained crowns (T2), 2 screw-retained abutments and screw-retained union crown (T3). The finite element analysis (FEA) was performed to explore the stress distribution of prosthetic restoration. Further linear regression analysis was performed to study the effect of crown height (CH) on the maximum stress of the implant system. The FEA results demonstrated that T3 restorations model exhibited the lowest level of stress. The overall peak stress of the implant-abutment assembly in T3 model were reduced by 69.6%-92.9% compared with T1 and T2 models. Furthermore, increasing CH values will lead to a strong linear increase in the maximum equivalent stress of the implant-abutment assembly. In terms of biomechanical property, the screw-retained union crown design for the double implant-supported prostheses achieved the best load resistance and optimal stress distribution compared to the other two single crown designs. Besides, screw retained union crown with smaller CH value was preferred for implant-supported restorations.

In this paper, we propose a set of effective channel combinations for detecting sleep arousal based on respiration-related channels and pulse-related channels based on PSG signals and deep learning methods. Different from the single or combined channels of electroencephalogram and electromyography used in most studies related to arousal detection algorithms, this paper inputs four signal channels of thoracic respiratory effort (THO), abdominal respiratory effort (ABD), pulse wave (Pleth) and pulse wave oxygen saturation (SpO2) from clinical PSG data into the neural network classification model, and achieves on the test set with precision 83.34%, recall 86.45%, specificity 99.16% and accuracy 98.57%. The proposed combination of signal channels for arousal detection is in line with the theoretical basis and is an innovative combination of channels, which is also at the leading position of detection performance compared with previous studies and provides ideas for future arousal detection using lightweight wearable devices. This paper is innovative, applicable and generalizable.

In recent years, combination pharmacotherapy has emerged as the standard clinical approach for treating a multitude of complex diseases. Nevertheless, the concurrent use of multiple drugs can lead to interactions that provoke adverse effects in patients, rendering the accurate prediction of potential drug interactions a critical task. Current drug-drug interaction databases are limited in scale, which often results in models learned from these sources struggling to generalization, particularly in cold-start situations where achieving satisfactory predictive performance is challenging. To surmount this obstacle, we propose a drug-drug interaction prediction method based on pretrained models, called PMSE-DDI. We design a pretraining strategy capable of simultaneously learning encoders for both molecules and substructures and use it pretrain on roughly one million unlabeled molecular. Thereafter, the pretrained encoders were transferred to the drug-drug interaction prediction model, and a substructure attention module was developed to pinpoint key substructures within drug molecules. Experimental results in cold-start scenarios demonstrate the superiority of PMSE-DDI in relation to existing methods, as well as the efficacy of the pre-training strategy.

With the promotion of the "new agricultural science" and the development of agricultural information technology, the animal science profession is developing rapidly, and many workers in the field of animal science are asking questions and acquiring knowledge through the Internet. Named entity recognition is a core fundamental technology in the field of natural language processing, which can identify entities and obtain useful information from various types of unstructured question and answer data, and then build question and answer systems, knowledge graphs and other applications, and is used by workers in the field of animal science. Although named entity recognition has been applied in many fields of Chinese, many Chinese characters have multiple meanings, and the word vectors obtained by traditional word embedding techniques are unable to show such multiple meanings of words.

Clustered regularly interspaced short palindromic repeats (CRISPR) / CRISPR-associated nuclease 9 (Cas9), as a powerful gene editing tool, makes remarkable development in genomics. Because it has several advantages such as easy-to-operate, accurate targeting and efficient editing, CRISPR/Cas9 has been widely studied in fundamental biomedicine and preclinical experiments. CRISPR/Cas9 system, designed from a prokaryotic CRISPR immune system, uses single-guided RNA (sgRNA)/Cas9 nuclease complex to bind target DNA for gene editing. Cas9 proteins cause double strand DNA breaks (DSBs), and DSBs are repaired by non-homologous end joining (NHEJ) and homology directed repair (HDR). Currently, it has showed some success in treating human diseases such as monogenic diseases. Herein, we review mechanisms of CRISPR/Cas9 system and the findings of CRISPR/Cas9 system for the therapy of hematologic diseases (including sickle cell disease, leukemia as well as anemia) and cancer, in order to provide some references and ideas for these two diseases both in basic research and clinical therapy as far as possible.

Infusion pumps are commonly used clinical medical equipment which can precisely control the speed of infusion and deliver nutrients or medications to the patients. The infusion accuracy of infusion pumps seriously affects the life and health safety of patients, and infusion accuracy is the most important factor to assess the reliability of infusion pumps. Representative researches on the factors influencing the infusion accuracy of medical infusion pumps are summarized systematically. Influencing factors are divided into four aspects: external environmental factors, internal structure of pump body (mechanical accuracy), auxiliary consumables, and parameter settings. This paper details the current status of research in these four areas and provides an outlook on future research directions for quality control of infusion pumps.

National policies encourage industrial technology upgrading, the development of independent intellectual property rights for testing instruments, and support breakthroughs in key technologies for in vitro diagnostic instruments, devices and reagents. The deepening of China's ageing population will bring more demand for medical diagnosis. Based on the background of intelligent medical treatment and home medical treatment, this study is committed to using blockchain technology, Internet of things technology and wearable technology to build a portable ECG monitoring system, so as to realize the local storage and upload of users' ECG detection data to the cloud, and real-time monitoring and preservation of ECG data.

Nowadays, the value contained in big data has attracted more and more attention. How to efficiently analyze big data has become an important topic. Machine learning is one of the popular data analysis methods. However, this method is difficult to apply to the massive and continuously growing datasets in the big data environment. How to transform traditional machine learning algorithms to make them more suitable for big data environment is popular in research. This paper studies the multimodal medical data statistic under machine learning algorithm, and uses the decision tree algorithm in the machine learning algorithm to investigate and study the medical data statistics.

The purpose of this study was to investigate the value of wavelet features in machine learning radiomics and deep learning for predicting pathologic complete response (pCR) to neoadjuvant chemotherapy (NAC) in breast cancer (BC) patients. We retrospectively analyzed 199 BC patients from the Duke-Breast-Cancer-MRI dataset. First, five machine learning classification models were constructed based on the combination of texture features and wavelet features extracted from the tumor mask region for the prediction of NAC response. Then, a deep learning model based on wavelet transform (WCNN) with a classical backbone of VGG16 was proposed to effectively predict the pCR of NAC using dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) scans taken before NAC treatment. The prediction performance was evaluated by using the area under the receiver operating characteristic (ROC) curve (AUC). Results show that both radiomics and deep learning models with wavelet features perform much better than the models without wavelet features. In addition, the proposed WCNN model outperforms all machine learning-based radiomics and deep learning models built in this study, which indicates that the wavelet-based deep learning model has great potential to provide information for breast cancer patients planning to undergo NAC treatment.

Medical air system is widely used in hospitals. Parameters such as terminal pressure, terminal pressure drop, terminal flow, water content, and particle concentration can directly decide the quality of medical air system. At present, there is no relevant calibration specification，or method for medical air system in China. This article presents a novel calibration method for medical air system by determining corresponding environmental conditions, calibration standards, location of standards and calculation equations. The calibration results show that the calibration method is scientific and practical for metrological performance evaluation of medical air system.

In order to meet the accuracy and real-time requirements of epileptic seizure prediction, a prediction method based on Dynamic Bayesian Network (DBN) is proposed in this paper. After a pre-processing of EEG recordings, the permutation entropy features are calculated and selected to establish DBN model. The structure leaning and parameter learning of DBN are carried out respectively based on scoring-search algorithm and maximum likelihood estimation algorithm. Finally, seizure prediction is realized through probabilistic inference of DBN. Evaluated on CHB-MIT EEG dataset, the proposed prediction method achieved an average sensitivity of 91.4% and false alarm rate of 0.14/h, with an average prediction time of 35.64 min. Experiment results demonstrate the good performance of DBN in seizure prediction.

Heart disease is one of the world’s deadliest health problems. An efficient and accurate approach to early prevention and detection of heart disease is becoming critical. Machine learning (ML) technology has demonstrated the capability and effectiveness to assist decision making in many fields. ML techniques have also been applied in heart disease prediction, but a single ML model usually cannot achieve good performance. Therefore, to improve the performance of the prediction model, a hybrid XGBoost and logistic regression (XGBLR) using improved hybridization of differential evolution and monarch butterfly optimization (IMBODE) are proposed. XGBLR reduces the complexity of feature engineering and enhances the predictive performance of the model, and IMBODE strengthens the ability to optimize model hyperparameters to achieve better solutions. The evaluation results demonstrate that the XGBLR is advanced in the prediction of heart disease, with an accuracy of 96.72% on the Cleveland dataset.

Alzheimer's disease is a form of mental loss that causes memory loss, slow and gradual changes in brain function. Early diagnosis and prediction of Alzheimer's disease can delay the onset of the disease and possibly prolong the life span of patients, which has important scientific significance for the whole society. We used the multiple linear regression model and established a statistical model according to the obtained test results and past experience to reveal the evolution patterns of different categories of Alzheimer's disease over time. The corresponding variables were predicted and controlled. The regression coefficient was calculated, and the correlation coefficient was 0.82, thus demonstrating the accuracy of the model. Secondly, since the diagnosis and characteristic indicators of patients are contingent to a certain extent, this paper draws the time series images of relevant personnel through random extraction, compares the predicted results with the real results and tests the residual error, with an accuracy of 80.8%.

Automated lung nodule detection methods are challenged by the fact that pulmonary nodules often present multiple morphologies and scales and by the low differentiation of nodules from surrounding tissues. To address the above issues, we propose a pulmonary nodule detection method based on 3D multiscale and semantic context heterogeneity. The method uses dilated convolutions with different dilation rates to construct a Context Fusion Module (CFM) to extract pulmonary nodule multiscale fused features, thus enhancing the detection ability of non-significant and small nodules. The Channel Semantic Context Heterogeneity Model (CSHM) is proposed to address the problem of noise and redundant information in multi-scale fusion features, to suppress noise and redundant information from feature channel-level contextual relevance and redundancy, to enhance potential discriminative features, and to reduce the interference of useless information, so as to enhance the ability of nodule detection and localization. Experiments were conducted on a publicly available dataset (Luna Nodule Analysis 16, LUNA16) with a high sensitivity and an average sensitivity of 83.6% over eight false alarm numbers, which is better than the other method, validating the superiority of the proposed method.

Acute kidney injury (AKI) is a common critical syndrome with a high incidence rate and high mortality. The risk assessment and prediction of patients with AKI are of great significance for the timely intervention and treatment of patients with potential risks. Based on the large scale of clinical information of multi-center inpatients (46, 395 AKI inpatients vs. 534, 066 non-AKI inpatients), the distributed federal learning framework was applied to construct the AKI prediction models. Experimental results demonstrated that the proposed method achieved promising performance in AKI risk identification and prediction.

Bipolar disorder (BD), which also refers to manic-depressive disease, often synchronously develops accompanying mental disorder, endocrine, and long-term awful physical conditions and has a poor prognosis, making bipolar disorder a significant public health problem and a global disease burden. The disorder is highly heritable and has been shown to be genetically linked. Identifying and screening major molecular markers for therapeutic targeting has become a global concern. This paper aims to use a bioinformatics approach to screen genes included in the pathogenesis of bipolar affective disorder and to analyze their signaling pathways. In this study, mRNA expression profiles in the orbitofrontal cortex (GSE5389) and dorsolateral brain tissue (GSE5388) obtained from the GEO public database on the National Center for Biological Information (NCBI) website were employed in order to detect and analyze differentially expressed genes (DEGs) in the normal and disease groups adopting the limma package with the R language, followed by an application of Metascape database for GO and KEGG signaling pathway analysis and WGCNA package for GSE5389 for weighted gene co-expression network analysis in order to acquire the relevant modules as well as crucial genes relation. It was shown that patients with bipolar disorder, genes included in the development of neuronal cell development, cell agglutination, immune signaling, and glial cell regeneration are dysregulated.

Sepsis is a serious clinical syndrome that can be life-threatening. Effective fluid resuscitation is crucial for improving the outcomes of patients with sepsis. During fluid treatment of sepsis, timely achieving a negative fluid balance is associated with better outcomes. Inappropriate fluid administration can increase risk of death in patients with sepsis. However, there is still uncertainty about the best strategy for managing fluids in sepsis. In this study, we present a machine learning model based on Extreme Gradient Boosting (XGBoost) algorithm to predict the initial timing of de-escalation phase of fluid therapy after fluid resuscitation for patients with sepsis. We collected a total of 1,353 sepsis patient records from the MIMIC-IV database for our study. From each patient, we extracted 37 hourly features that reflect their status. To assess the importance of each feature to the timing prediction, we used the SHapley additive exPlanations (SHAP) method. Based on all 37 features, our full model demonstrated a precision of 82.02%, a recall of 85.99% and a specificity of 97.03% with an area under the ROC curve (AUC) of 0.9863 well demonstrating its potential in helping clinicians determine the optimal timing for fluid de-escalation in sepsis management.

ECG signal is an important basis for clinical diagnosis of cardiovascular disease, and ECG signal automatic detection of cardiovascular disease is the core of automatic diagnosis. In this paper, based on the existing ECG signal detection and location algorithm, a differential absolute value extremum algorithm is proposed. This algorithm combines the original difference algorithm principle and the difference absolute value, at the same time in the final positioning abandoned the discrete sampling location method, and using the extremum positioning method, which improves the positioning accuracy. Using this algorithm, the QRS waveforms of 5 groups of ECGs were detected and the detection accuracy was 99.67%.

The electrocardiogram (ECG) is a graph that records the electrical activity changes of the heart during each cardiac cycle. Therefore, it is of important clinical and medical significance to pay attention to ECG research. In this paper, we selected the appropriate parameters such as wavelet basis, threshold and the number of decomposition, and the different noises in the ECG signal were filtered by wavelet soft threshold transformation. Then, a 10 layer CNN model was used to classify and recognize the preprocessed ECG signals. Experiments on the MIT-BIH database verified that this model had a good classification effect.

The segmentation effect is not very good due to some problems such as inconsistent thickness of blood vessels and poor contrast of blood vessel boundaries. In order to solve the problem of small blood vessels breaking during network segmentation caused by the above reasons, this paper takes U-Net as the network basis, and firstly adds CBAM module in the process of network coding. Secondly, PPM module is introduced into the model to fuse the vascular features of different scales, so as to obtain more rich contextual information for the original. In addition, CDCM module is added at the bottom of the model to enhance the receptive field of the model, so that the network has more spatial scale information, and strengthen the awareness of context features, so that the segmentation task has better performance. The experiments on DRIVE, CHASE_DB1 and self-made dataset100 show that compared with the original U-Net, SE increases by 9.22%, SP by 0.83%, Acc by 1.65%, AUC by 1.56% and F1 by 5.93%. Compared with other methods, the sensitivity, accuracy and other indicators are improved, indicating that the vascular segmentation method in this paper has the ability to capture complex features and has higher advantages.

The decoding of electroencephalogram (EEG) signals plays an extremely important role in brain-computer interfaces (BCI). However, the processing of physiological signals, particularly the decoding of multi-channel EEG signals, still poses significant challenges. Past deep learning methods often relied on subject-dependent settings, which resulted in new users needing to perform complex calibration procedures before they could use BCI devices. Therefore, we proposed a novel end-to-end deep learning model, MRMHNet, for motor imagery (MI) classification. Firstly, we utilized a feature extraction block based on a Multi-Resolution convolutional neural network (MRCNN) to extract features in both frequency and spatial domains. Secondly, we utilized a block based on the Multi-Head Attention (MHA) to extract global temporal information of the features. Finally, we validated the classification performance of our method using OpenBMI datasets, and the results showed that our method achieved the highest accuracy in both subject-dependent and subject-independent settings. Specifically, in the subject-independent setting, our method achieved the highest accuracy and F1-score, with values of 73.74±13.35% and 73.33±14.87%, respectively. This indicates that our method has good classification performance and high practical value in the field of BCI.

In cases of prolonged sitting, cross-legged posture is commonly used. Cross-legged sitting's causes, benefits, and drawbacks have been studied in the general population, but not in adolescent idiopathic scoliosis (AIS) patients. To address this issue, the buttock pressure distributions of 11 healthy participants and 27 AIS patients were acquired by piezo-resistive sensor array. Pressure parameters, asymmetry index and equilibrium parameters were compared between upright sitting and cross-legged sitting. The results demonstrated that crossing legs promoted relaxation but could also result in high-pressure regions, pelvic deformities and unstable seated postures. These effects were more significant in the AIS group, indicating that AIS patients could receive a higher degree of harm when crossing their legs. According to our research, cross-legged sitting should be avoided, due it can have both short- and long-term negative effects.

The depth of anesthesia is an important indicator for determining the clinical dosage of anesthesia. Currently available monitoring methods have problems with accuracy relying on human experience for judgment and poor precision. Electroencephalogram (EEG) signals have the characteristic of reflecting the mental state of the cerebral cortex. Therefore, this paper proposes a feature extraction method combining EEG signal symbolic entropy and short-time Fourier transform, as well as an anesthesia depth monitoring method based on least squares-support vector machines (LS-SVM) classification. In addition, this paper uses the anesthesia depth measurement standard based on the energy ratio of various rhythmic signals to divide the anesthesia depth into four levels. An anesthesia depth monitoring experiment was conducted using a self-built EEG signal collection platform. According to the experimental results, the proposed method can accurately classify the depth of anesthesia, with a classification accuracy of 86.87%.

Aiming at the poor experience of home-oriented sleep monitoring technology and the insufficient accuracy of single physiological signal for sleep staging, this paper proposes a sleep staging method by cardiopulmonary signal. Based on the non-contact high-sensitivity optical fiber sensor that satisfies high comfort at home, high-quality cardiopulmonary signal during sleep is obtained, and through algorithm analysis of cardiopulmonary signal separation-extraction synchronization, it overcomes the interference of body motion and realizes a high-quality A sleep staging algorithm for cardiopulmonary signal. Comparing with polysomnography for whole night sleep, this method can achieve an overall accuracy of 83.25%, providing a solution for sleep staging monitoring in a home environment.

Sleep is a very important physiological process for humans, and sleep staging is an important basis for evaluating sleep quality and diagnosing sleep-related diseases. Manual sleep staging is time-consuming and laborious, and the staging results are easily affected by subjective factors. Therefore, the research on automatic sleep staging models has extremely high research value and clinical application value. This paper proposes a sleep staging method based on multi-scale one-dimensional convolutional neural network and self-attention mechanism. The attention mechanism mines the channel correlation and temporal correlation of different channels respectively. This paper selects the sleep data of 153 normal subjects in the public dataset Sleep -EDF for model training and testing. The average staging accuracy rate was 84.5%, and the consistency test value was 0.781, which was better than the consistency test value of 0.75 for the inter-physician score. The experimental results show that the algorithm model in this paper has a high staging effect, can make full use of sleep data, and has important research significance for assisting clinicians in rapid sleep staging.

Indoor dust mites are one of the common and major inhalant allergens, and detection of specific antibodies against dust mite allergens can help in the prevention and treatment of allergic diseases and in the study of immunotherapy. However, traditional assays such as enzyme-linked immunosorbent assays are time-consuming and costly, and are not conducive to rapid detection. Microfluidic chips enable precise control and manipulation of tiny fluids, allowing for rapid, low-cost, high-throughput bioassays. Digital enzyme-linked immunosorbent assays have the advantages of high sensitivity and specificity. In this paper, fluorescence-encoded micro-magnetic beads were used as solid-phase carriers to modify Der p1 dust mite antigen, and then combined with patient serum and detection antibody to form an immunosandwich structure. Then, we successfully detected Der p1 dust mite allergen-specific antibodies in patient sera using a combination of microfluidic chip and digital ELISA, and plotted the standard curve. The method has promising applications and can provide an essential reference for clinical diagnosis and treatment.

Currently, a large number of studies analyzing the impact of genes and food intake on hypertension have been conducted using genomics, metabolomics, and multi-omics. However, when it comes to human activity, machine learning models face difficulty in adapting to the dimension, length, and complex activity combinations of exercise data. As a result, current research only focuses on quantitatively analyzing specific exercise behaviors. There is a lack of research on using deep learning methods to learn the relationship between hypertension and exercise from accelerometer data. We propose an optimal sequence mapping algorithm based on a machine learning model that uses a hidden semi-Markov model to downscale and encodes accelerometer data. We then propose an activity combination module to characterize complex activities. Finally, we use this method to train a random forest classification model and evaluate our experimental results. Our experimental results demonstrate that our improved sequence mapping algorithm has significantly reduced the length of effective wear time for accelerometer data by approximately 64.77%, rendering it well-suited for use with DL models. Furthermore, our findings show that the RF model with a module size of 3 outperformed both single-activity modules and other activity combination modules. Our research demonstrates that the HSMM improved sequence mapping algorithm enables personalized selection of the optimal activity threshold. Our encoding method effectively reduces the length of accelerometer data sequences while preserving key features. Moreover, our activity combination module accurately characterizes complex human activities, significantly enhancing the predictive power of the RF model.

This paper evaluated the cuffless and continuous blood pressure measurement performance of single-channel photoplethysmography (PPG) signals using different end-to-end deep learning models on the same dataset. The proposed a signal quality detection algorithm to effectively remove unusable signal segments in PPG signals. Popular end-to-end deep learning models were validated on various blood pressure and PPG signals obtained from the intensive care unit database. The experimental results showed that InceptionV1's multi-scale convolutional network achieved the best performance.

This paper presents a low-power, low-noise neural recording dual-threshold adaptive action potential detector, a circuit that provides a method for detecting action potential signals with low amplitude to reduce the false and missed detection rate of action potentials, adaptively generates two suitable comparison thresholds - peak and trough thresholds and the period of the action potential can be detected. Simulation results show that the system achieves an input reference noise of 14µVrms, an IF gain of 90dB and a power consumption of 54µW. The proposed low-power, low-noise dual-threshold adaptive action potential detection circuit is suitable for implantable devices with neural recording systems.

Monitoring and recognizing sleep breathing events have important implications in the preliminary diagnosis of respiratory diseases. However, conventional sleep breathing monitoring devices that come in contact with human skin can cause sleep discomfort and thus affect sleep outcome, and they cannot monitor sleep breathing events for a long time. In this study, we propose a non-contact system based on fiber optic sensing for long-term sleep breathing event monitoring. A K-nearest neighbor(KNN) respiratory event classification method based on Gaussian kernel function and ball tree search is proposed to improve the accuracy of respiratory event classification and apnea localization. The changes in morphological features of respiratory signals are characterized by time domain, frequency domain, and waveform decomposition for feature extraction. The system collected overnight sleep breathing data from 30 subjects in sync with a polysomnography system, totaling 2000 breathing event samples (100 s per sample), and tested the enhanced KNN classification model. The experimental results show that the accuracy of the classification model reaches 96.4%, which is 2.1% higher than that of the traditional KNN classification model. The training time is reduced by 1.71 s, and the testing time is reduced by 0.02 s.

Traditional networks cannot focus on which features are important when extracting features and use the same weighting for all features, so this paper investigates breast lump detection based on attention mechanisms. By adding three attention mechanisms, SE, ECA and CBAM, respectively, to the backbone network of YOLOv5s, the network's ability to extract breast lump features is enhanced for breast lump recognition. The experimental results show that adding SE or ECA or CBAM attention mechanisms alone to the YOLOv5s network, they are higher in accuracy, recall and mAP@0.5 than the original YOLOv5s network model, and the model with the addition of CBAM attention mechanism improves the accuracy, recall and mAP@0.5 by 3.1%, 2.6% and 3.1% respectively, which is better than the models with the addition of SE and ECA attention mechanisms. The proposed attention mechanism-based breast lump detection model can effectively detect breast lumps and provide help for early screening.

A multifunction, adjustable frequency and current intensity transcranial pulse intelligent monitoring stimulator is researched and designed in this paper, which takes STM32 chip as the control core, realizes stable pulse electrical signal output through pulse signal shaping module and fuzzy PI constant current control, embedded human physiological parameter monitoring module can monitor human body physiological parameters such as ECG, blood pressure, blood oxygen and body temperature in real time, and constructs upper computer module. Real-time display of monitoring data and adjustment of the size of the pulse electrical stimulation output current amplitude, the stimulator stimulation has high precision, strong applicability, convenient parameter adjustment, etc., and has a certain practical value for clinical rehabilitation treatment.

Cardiovascular diseases are the most common type of diseases with the highest mortality rate, and the most common of these diseases is myocardial infarction. Regarding the prediction of myocardial infarction, the traditional method is judged by the expertise and experience of medical personnel, but this method may lead to inaccurate results due to differences in judgment criteria, while using deep learning methods to study myocardial infarction facilitates the processing of features and can improve the accuracy of diagnosing the disease. In this paper, we propose a CNN_SVM model of convolutional neural network combined with support vector machine to predict myocardial infarction. A total of 87 features are obtained based on the vital signs and biochemical examinations of patients in the MIMIC data set. For these features, the CNN_SVM model firstly performs PCA dimensionality reduction by mapping the data from high dimension to low dimension to reduce the dimensionality of the features and retain as much information as possible to complete the dimensionality reduction, and finally obtains 45 data features; secondly, the last layer of the CNN model, Softmax classification layer, is replaced with SVM classification, and the final results are two kinds, namely myocardial infarction and non-myocardial infarction, respectively. The model construction was completed. The experiments showed that the accuracy of CNN_SVM model was 96.67%: the accuracy was improved by 7.60%, 7.15% and 17% when compared with CNN, XGBoost and MLP, respectively.

This study investigated an improved method for predicting brain activity using 7T fMRI data. The data resolution was 7T, and the stimulus and fMRI measurement data were paired. Text descriptions were used as original features and were deep learning encoded for model training. The results demonstrated that our model effectively predicted brain activity response, surpassing some previous methods. During experiments, increasing the encoding dimension improved the model's fitting performance. Our study enhances the ability to simulate the brain and investigate its cognitive processes.

As an important restoration substance in cells, reduced glutathione (GSH) plays an important physiological role in the human body. In this paper, GSH was determined by indirect Spectrophotometry. Through the optimization experimental conditions, the amount of Fe3+-1,10-phenanthroline chromogenic concentration was 0.8 mg/mL, and the optimum pH value was 4.0 mL, and the amount of buffer solution was 5.0 mL, and the heating time and cooling time were 10 min. In addition, the absorbance was measured at 510 nm, and GSH was in the range of 0~0.4 mg/mL. The linear regression equation was A=1.7981c+0.1348 (which A represents the absorbance and c represents the concentration of GSH), and the correlation coefficient was r2 = 0.9955. The method was applied to the determination of reduced GSH in three GSH samples. The indirect spectrophotometry method is simple, low cost and easy to operate.

At present, one-stop CTA is considered as the gold standard in the preoperative evaluation of TAVI. It analyzes and measures the anatomical structure of the aortic root to match the best artificial aortic valve, which is very helpful for the smooth progress of TAVI surgery. Compared with Echocardiography, CTA has a higher prediction accuracy.

Traditional deep learning based medical image segmentation methods require a large number of labeled datasets, while medical images lack datasets due to various reasons such as patient privacy, and even more so lack labeled datasets. Therefore, this paper proposes a medical image segmentation method based on cycle consistency data augmentation. First, train space transformation registration network and appearance transformation registration network based on cycle consistency respectively. Then, use the trained space transformation registration network and appearance transformation registration network to perform data enhancement on the original dataset. Finally, the augmented data is used to train segmentation network to achieve medical image segmentation. Experiments on public datasets show that this method has better segmentation effect and better image registration effect.

Accurate identification and removal of the tumor is critical for effective treatment of brain tumors. Manual identification of the tumor location from medical images can be challenging. Hence, modern medicine relies increasingly on artificial intelligence models to assist with segmentation. Despite this, currently available medical image segmentation models still do not produce satisfactory results in segmenting lesions like tumors. Many times, medical data is not sufficient for models to learn effectively, making it difficult to accurately segment the target. Therefore, we propose a new model, E-U-Net++, it combines data augmentation methods with existing U-Net++ series models enabling better segmentation results even in situations where medical data is generally scarce. In the training phase, before inputting the original training data to the U-Net++ model, we incorporated various data augmentation techniques, including random flipping, cropping, and the addition of noise. During the testing phase, the preprocessed data was subjected to center cropping and then provided to trained U-Net++ weight for processing. Experimental tests comparing the testing consequences of E-U-Net++ with the U-Net and U-Net++ algorithm revealed that the improved E-U-Net++ algorithm outperformed these models in brain tumor segmentation. Enhancing the original segmentation model through data augmentation methods holds great potential in improving brain tumor segmentation performance.

The study on the structural connectivity and functional connectivity of the brain networks of mathematically gifted adolescents found that mathematically gifted brains exhibited cognitive neural features that are different from average adolescents and beneficial for task processing in the aspects of activation in task-related brain areas, information interaction between hemispheres, and network connection reorganization. In this paper, we adopted directed transfer function (DTF) and partial directed coherence (PDC) to analyze the effective connectivity of mathematically gifted adolescents from electroencephalography (EEG) data. The information flow patterns of the brain networks during the deductive reasoning task were analyzed and the group differences in the information flow of the fronto-parietal network were compared at the θ-band and γ-band. The empirical results demonstrated that the mathematically gifted brain exhibited more information transfer between distant nodes and more information flow from the prefrontal lobe to other brain regions in both DTF and PDC directed networks, ensuring the excellent task performance of mathematically gifted adolescents.

This research was proposed in the background of the rapid development of artificial intelligence technology. The traditional medical device has a low level of intelligence, mainly relies on manpower for control, consumes a lot of manpower, and the function is relatively simple. With the development of artificial intelligence technology, it has promoted the fusion of artificial intelligence and medical devices R & D. This is an important product of the development of artificial intelligence technology. In view of the shortage of beds that exist when patients receive drip injections, a large number of medical staff are needed, the change of medication may not be timely, the drip time is long, the patient is bored, and the current medical environment needs to be improved, etc., this paper designs a drip assistant intelligent medical robot based on artificial intelligence technology. This research is of great significance to the in-depth research and application of artificial intelligence technology in the medical field.

The robust performance of machine learning techniques provides good application prospects for neuroimaging, not only breaking the limitations of human eye recognition, but also improving the generalization ability of image features. By introducing machine learning and neuroimaging into the field of acupuncture, the " acupuncture-point-disease" target advantage can be more objectively and visually exploited, making it efficient for clinical use. By synthesizing the relevant literature, the author takes multidisciplinary intersection as a starting point and machine learning combined with neuroimaging as a means to sort out the acupuncture point specificity, acupuncture modality and acupuncture efficacy prediction. And we also review the progress of the current research on "bottlenecks" and future research trends.

We create a randomized age-structured HIV/AIDS framework in this paper which that takes into account both ART therapy and outside noise. We explore the model's positive solution's existence, uniqueness, and boundedness to guarantee its validity. We also offer error estimates and suggest a completely discrete approach employing the Euler scheme for time discretization and Galerkin method for age partition. Finally, numerical simulations were carried out after feed-forward neural networks and applicable data from the Chinese Centre for Disease Control and Prevention were used to derive expressions for the model parameters.

In this study, we constructed a mental stress recognition model with Electrocardiogram (ECG) signals by machine learning. Firstly, we collected the ECG signals while students are explaining simple math exercises, located the R-wave peak of the QRS wave group, and calculated the R-wave to R-wave (RR) interval. Then we extracted the characteristic parameters of autonomic nervous response from the RR interval, carried out statistical analysis and sequential forward selection and finally constructed a psychological stress recognition model. The accuracy of the model in the test set and the independent subject validation set was 79% and 83%, respectively. The results show that it is feasible to recognize strong or weak psychological stress state through machine learning method.

Behavior recognition is one of the most popular fields in computer vision, and graph convolution network for skeleton based data has been widely used in the field of action recognition, and has achieved remarkable results. Most of the previous studies focused on modeling the relationship between adjacent joint points in natural state, while ignoring the graph topology between non adjacent joint points. In this work, we propose a new graph convolution with EM dynamic routing to learn different structural features dynamically in action recognition based on human skeleton data, and cluster multiple joint points into corresponding graph topologies effectively. The proposed EMD-GC takes the initially artificially defined shared topology as the general prior knowledge of the model, then models the topology according to the specific correlation of each channel, and finally clusters the features into the corresponding topology through the GMM model. Experiments show that EMD-GCN is superior to the most advanced methods on NTU RGB+D dataset.

At present, the diagnosis of schizophrenia mainly depends on scale evaluation and expert interview, which is subjective. It may lead to misdiagnosis. Therefore, an objective classification algorithm of schizophrenia based on multimodal information fusion is proposed. First, the brain magnetic resonance images (MRI) of schizophrenic patients and healthy controls were acquired. Then, information from multi-modal MRI was extracted and fused to build a graph. At last, a deep graph neural network was used to classify the individual graph of two groups GradCAM method was used to visualize the feature weights of the model. It’s found that two groups of participants can be effectively classified based on graph features. It also revealed that neuroimaging features from the occipital lobe, frontal lobe and limbic system played an important role in classification process. The present work indicated that the integration of neural science and information science has good application potential in disease diagnosis and pathological imaging.

In view of the problem of inaccurate outpatient infrared temperature measurement, this paper expounds the principle of infrared temperature measurement, obtains the influence of influencing factors such as emissivity, distance, and ambient temperature on the accuracy of human forehead temperature measurement through experimental research. And the precise technical strategy of outpatient infrared temperature measurement under epidemic prevention and control is put forward, pointing out that the core is to strengthen the training of personnel in the use of instruments, so that they can understand the basic working principles and correctly set the relevant parameters of infrared temperature measuring instruments. It is further proposed that during the high incidence of the epidemic in autumn and winter, the alarm temperature value of the infrared thermal imager should be appropriately adjusted for fever patient screening. Through the investigation and research method, the sensitivity of infrared thermal imager to accurately detect and screen fever patients has been greatly improved compared with the general detection method, from 82.2% to 95.9%. The results of this paper can enable outpatient medical workers to understand the basic principles of infrared temperature measurement, and can effectively guide the outpatient infrared temperature screening.

The default network has been one of the hot topics of research in cognitive neuroscience in recent years, and studies have reported that it may be involved in a variety of cognitive activities and that certain psychiatric disorders are associated with its abnormal activities. However, the directional connectivity relationships (effective connectivity patterns) between major brain regions within the default network are still unclear. In this study, the effective connectivity patterns between four core brain regions of the default network were explored using an international cutting-edge spectral dynamic causal modeling algorithm. The experimental results revealed that the posterior cingulate gyrus within the default network receives information input from the medial prefrontal and bilateral inferior parietal lobes, and may play the role of an information aggregation center; while the bilateral inferior parietal lobes have information input to both the medial prefrontal, and posterior cingulate gyrus, and may play the function of information drive and regulation within the default network. We believe that the changes in effective connectivity of nodes within the DMN accompanied by changes in information transmission are helpful for our deeper understanding of the function of the default network and have potential reference applications for the study of related psychiatric disorders.

Combining multi-site datasets is very important for increasing statistical power. With the development of medical imaging technology, the acquisition of MRI neuroimaging datasets with larger sample sizes has become possible. However, different scanners will generate non-biological noise in MRI neuroimaging data, leading to inconsistent research results across different scanners. In this research, we proposed the Auto-fader network to harmonize MRI neuroimaging data among different scanners to address this issue. The Fader network is a domain adaptation technique, it has an encoder-decoder architecture that can obtain the features without site effects, but the site labels must be encoded manually during training. Encoding site labels manually is only sometimes accurate. Therefore, on this basis, the Auto fader model has added site effects auto-encoding module and directly uses 3D convolution layers to extract features from the MRI neuroimaging data. Finally, experiments demonstrate that the Auto-fader model outperforms existing approaches in harmonizing MRI neuroimaging data from different scanners.

This article demonstrates the background information about the surgical soft robot at the beginning, showing clearly the developing history of surgical soft robots. Then the article discusses and explains the material, actuation, sensing and manipulation of the surgical soft robot in depth. What’s more, the specific applications of minimally invasive surgery are demonstrated, as well. At last, there are some challenges and problems reflecting that we still have a long road to go in designing better and better surgical soft robots.

In order to maximize the utilization effect of drugs in aerosol inhalation therapy and achieve "targeted" drug delivery, the transport and deposition characteristics of aerosol particles in the upper respiratory tract under different respiratory conditions were studied. Firstly, a 3D image of the upper respiratory tract of an adult male was taken by CT, and the 3D image data was reconstructed by 3D Slicer software to establish a real mouth and throat model of the upper respiratory tract model. The Langer method was used to discretize the particles. Finally, on the FLUENT platform, the numerical simulation of the aerosol inhalation process under different particle sizes and different breathing flows was carried out by using large eddy simulation (LES). The results show that: in the upper respiratory tract, the flow state of the airflow is basically turbulent, and the airflow structure in the upper respiratory tract is complex, and the airflow velocity changes sharply at the entrance of the mouth, the pharynx, the larynx and the transition of the trachea. Aerosol particle size decreases, particle deposition efficiency increases, and increased respiratory flow also leads to increased deposition efficiency.

Objectives: To provide reference for the evaluation and diagnosis of chronic neck pain (CNP) syndrome. Methods: 22 patients with CNP and 22 healthy people were recruited from South China Normal University, who were all college students. The static electromyography (EMG) of the subjects’splenius capitis muscles and upper trapezius muscles was detected by a surface EMG device during isometric contractions of muscles, and the dynamic EMG of the subjects’ splenius capitis muscles and upper trapezius muscles was detected during head and neck flexion and extension. Results: (1) There were very significant differences in the average EMG of the neck muscle, which was isometrically contracted, between patients with CNP and healthy subjects in the head neutral position, head natural forward position, and head maximum flexion position (P < 0.001). (2) There were significant differences in the integral electromyography of the neck muscle, which was isometrically contracted, between patients with CNP and healthy subjects in the head neutral position, head natural forward position, and head maximum flexion position (P < 0.005). Conclusions: (1) Both splenius capitis muscles and trapezius muscles of college students with CNP had functional damage, and the functional damage of the former was severer. (2) During head and neck flexion and extension, FER of neck extensor muscle EMG in college students with CNP significantly increased, indicating that the active activity functions of neck extensor muscles of college students with CNP were insufficient and the flexion-relaxation responses were weakened.

By using the B3LYP method of density functional theory (DFT), the nitrogen heterocyclic carbene palladium complex was theoretically calculated and simulated by ECP28MDF and 6-311G* base group, respectively. And after the structure was optimized, the infrared spectrum, molecular binding energy, UV-visible absorption spectrum and frontier molecular orbital energy level of the optimized N-heterocyclic carbene palladium were simulated and analyzed. The results showed that the palladium ion and n-heterocyclic carbene palladium were well combined, and the stability of the complex formed was good, which means that n-heterocyclic carbene palladium has strong trapping ability and binding energy.

In recent years, segmentation of the multimodal brain tumor image puts forward high requirements for performance. To meet the accuracy requirements, we propose a multimodal brain tumor image segmentation method based on UNet and deep residual learning, which can make full use of multimodal information in order to obtain details. The model achieves the information extraction and integration of multimodal images at different scales by introducing a multiscale feature extraction module, a coordinate attention module, and a dense atrous spatial pyramid pooling module. The results show that the proposed method can fully exploit the multi-scale context information of multimodal images, alleviate the problem of blurred edges in segmentation regions, and achieve good performance on popular evaluation metrics.

Respiratory tract infection is a major disease that affects children's physical development and physical and mental health. Vitamin D intake has a certain internal relationship and correlation with the incidence of respiratory tract diseases. We selected children with respiratory tract infection as the research data to study the correlation between vitamin D intake and respiratory tract infection in children. In order to better analyze data information, multi-view clustering as an important unsupervised learning method is applied to the above research. In this paper, we propose a novel partition-level multi-view clustering method with weighted adaptive graph learning, which not only captures global and local structures simultaneously, but also fuses multi-view information through a weighted fusion mechanism. The effectiveness of our proposed method was verified on three public datasets, and the method was applied to respiratory disease datasets to achieve good clustering results.

For a long time, classification in dynamic videos on public datasets have been a hot research topic. However, in medical image processing, many methods still use traditional machine learning methods to address the problem of cellular dynamics classification. In this study, we paid more attention in the morphological changes of lymphocytes in healthy mice after skin transplantation. We applied a typical deep learning network Slow-Fast into cellular dynamics classification task, in which we incorporated data augmentation techniques such as rotation and reverse ordering to alleviate the overfitting problem to a great extent. Our method achieved better performance on the live-cell video dataset compared to previous approaches.

The development of artificial intelligence has promoted the improvement of medical and health technology. In addition, in the nursing industry, artificial intelligence can effectively respond to patient situations and timely alert medical personnel. In nephrology, patients often have blood and urine related problems, and the problems usually come suddenly. Therefore, this article used artificial intelligence technology to design a management model for the risk of renal medicine nursing. This article mainly used qualitative and quantitative analysis and fuzzy theory to explore the relevant risk management models of renal medicine nursing. Experimental data showed that the coefficient of change for each indicator of the nursing staff factor in risk management was distributed between 0 and 0.14, indicating high concentration and low dispersion in the distribution.

Skull defects are generally caused by strong blows to the head, and surgical treatment is the preferred therapy for cranial injuries. With the development of computerized three-dimensional reconstruction technology and the improvement of related software development, the use of rapid prototyping technology combined with biologically active artificial bone materials can produce skull models that highly match the patient's skull defect site and promote good bone fusion. Herein, based on CT images of patients with skull defects, combined with information about the patient's brain tissue to obtain the complete original outline of the skull and restore the information of the skull defect site. After 3D reconstruction, a customized rapid prototype manufacturing of the skull defect repair implant is performed using a 3D printer. The resulting skull defect repair implant is very suitable for the site of the skull defect, indicating that the precise manufacture of skull defect repair implants using rapid prototyping technology is feasible, and has important significance for preoperative planning and reducing intraoperative harm during skull defect repair surgery.

Various factors caused fatigue in driving and were difficult to determine by one aspect. Most of the current research was focused on the recognition of facial features under the fatigue state, ignoring the fact that fatigue is a slow process, which cannot be identified quickly and requires multi-angle monitoring to determine fatigue, and that little research has been conducted combining both facial and physiological states. Because of the intelligence of vehicle systems and the emergence of intelligence watches with physiological parameters, it is impossible to monitor fatigue from multiple angles and with less impact on the driver. In this paper, based on face tracking, Percentage of Eyelid Closure Over The Pupil Over Time(PERCLOS), Eye Aspect Ratio(EAR), Heart Rate Variability(HRV), Blood Oxygen Saturation(SO2), and Electrocardiogram(ECG). Based on a 2-hour highway experiment, facial features and physiological states were collected from three drivers driving for 12 hours. Finally, it was concluded that the recognition method of facial features and physiological states is superior to the single face recognition method. At the same time, the reference to time and frequency domain HRV parameters will detect fatigue in advance.

Respiratory rate in vital sign monitoring is an indicator of great concern, and the realization of home vital sign health monitoring is importantly helpful in the early warning of respiratory diseases. When performing respiratory signal monitoring, the presence of interference signals such as BCG signal, body motion signal and environmental noise reduces the efficiency of respiratory signal monitoring. In order to improve the efficiency of respiratory signals, this paper proposes a multi-area cooperative vital sign acquisition system based on fiber-optic micro-vibration sensing and a respiratory signal quality assessment method. For the acquired raw signal, the respiratory signal is first obtained by zero-phase filter, and the respiratory signal with minimum pseudo-peak interference and consistent waveform amplitude is found by the positive predictive value and amplitude consistency, which provides a reference for the signal quality assessment of the respiratory signal.

Thoracic aortic aneurysm (TAA) is a prevalent disease that can cause considerable mortality if the aneurysm ruptures. In recent years, computer has played an increasingly irreplaceable role in the researches on the rupture risk prediction of aortic aneurysm. However, computational numerical studies on the influencing factors of TAA distensibility indicated by wall compliance and stiffness have rarely been addressed in literatures. In this work, CT-image based patient-specific TAA numerical models were constructed by computer. Fluid-solid interaction was considered in the simulation. The variation of TAA distensibility was observed as the mechanism of wall material behaved differently. Results also showed that the release of wall boundary constraint could lead to a decline in TAA compliance and an increase in TAA stiffness. Moreover, study found that different sites of the aneurysm had different distensibility. Quantitative study on the TAA distensibility was presented in this paper. It may provide a new basis on the rupture potential of TAA. This study may have important implications on future patient-specific diagnosis and intervention for individual TAA patients.

At present, more and more studies prove that physical activity (PA) has a positive impact on the health of the seniors and emphasize the importance of personalized exercise prescription according to the specific situation of the seniors. However, there is a lack of insights and consensus on PA, its influencing factors and design of the personalized exercise system to promote PA behavior of the seniors, especially people in retirement. The purpose of this study is to explore the design of the personalized intelligent exercise system for the seniors from the perspective of behavior design. Firstly, an online investigation of the seniors (≥50 years old) exercise was completed based on the theory of Behavior Design. Secondly, the relationship between PA behavior indicators and PA behavior of the seniors was analyzed and the PA behavior model was developed. Finally, the frame of the personalized intelligent exercise system “PIES” for the seniors was designed. The results show that PA behavior of the seniors is related to the indicators of motivation, ability, prompt, and individual characteristics, consistent with the Behavior Design basically. It is feasible to design the personalized intelligent exercise system based on Behavior Design. This study provides a reference for the design of the personalized intelligent exercise system for the seniors at the level of design framework. In the future, the specific content of the system will be further studied.

In this study, we developed a nasal endoscopy simulation system based on Unity to be used for training novice doctors. The nasal cavity and internal structures were segmented from high-resolution CT images of patients, and mesh models were generated. To address model penetration between surgical tools and the nasal cavity model, we proposed a collision detection method based on motion trend prediction. By predicting potential collisions based on motion trends and performing corresponding preprocessing, we were able to prevent model penetration. The deformation of soft tissue structures within the nasal cavity was achieved by using a voxel-based PBD model. By voxelizing the mesh model and adding PBD constraints to the voxelized model, we achieved synchronous deformation between the mesh and voxelized models through coupling. For haptic rendering, two Phantom Omni devices were used to operate endoscopes and other surgical tools while providing force feedback. Experimental results showed that our system can provide realistic nasal endoscopy scenarios and effectively assist doctors in surgical training.

Knowledge graph is a kind of knowledge representation, which captures information about entities, entity attributes and relationships between entities in a structured way, and is widely used in intelligent retrieval, recommendation systems, intelligent question answering, etc. Knowledge map is a graphical representation of the relationship between different concepts and topics in a specific field, while BERT is a most advanced language model that can understand the context and meaning of words in text. By combining the two methods of medical knowledge graph (MKG) and the bidirectional encoder representation of BERT model, it shows hope in improving medical information retrieval and decision-making, and can create a more comprehensive and accurate representation of medical knowledge, which can be used to guide clinical decision-making and improve the prognosis of patients, and ultimately improve the effect of BERT in the medical field.

Oral obstructive teeth are abnormal tooth development due to dysplasia or pressure on surrounding teeth and usually need to be removed during oral surgery. Prior to surgery, the surgeon needs to perform diagnostic imaging to determine the location and shape of the blocked tooth and to plan the surgical plan. Traditional diagnostic imaging methods mainly use radiographs and CT images, but these methods require manual measurement and analysis, which are complex and prone to errors. There is a fast development in Computer Vision and Deep Learning, semantic segmentation recognition of oral obstructive tooth images has become a new solution. In this paper, we will study the semantic segmentation recognition of oral obstructive dental images and propose a deep learning based semantic segmentation method for oral obstructive dental images.

This paper proposes a design of intelligent energy-saving lighting control system based on ZigBee and NB-IoT technology. The system uses low-cost ZigBee networking scheme for data communication within the streetlamp group, collects data to the gateway configured with NB IoT module, designs communication protocol for streetlamp application scenarios, and uses it to interact with cloud platform for data, so as to solve the problems of untimely control, vulnerability to interference and small coverage of previous communication technology. The stability of ZigBee network transmission is optimized, And design and implementation of information transmission data to the client. The test results show that the intelligent energy-saving lighting control system based on ZigBee wireless network technology and NB-IoT remote communication technology has the advantages of strong penetration ability, low power consumption, stable and reliable operation, and has strong practicability and promotional value.

The development of deep learning has significantly altered computer vision and given rise to a wide range of pre-made training models. It is currently utilized often in medical pathology imaging. Histopathology image analysis with computer assistance can speed up diagnosis and increase precision. This study aims to address the issue of classifying images of central nervous system tumor pathology using a new deep learning network architecture (SPACH). Ultimately, our model had a classification accuracy of 92.5%.

An important tool for effectively preventing blindness from eye diseases was an early screening of ophthalmic diseases. Eye fundus image scanning was the most cost-effective and rapid way to screen for ophthalmic diseases. The lack of medical resources and the inefficiency of diagnosis have been effectively improved using deep learning in eye fundus image screening. However, the detection effectiveness of the model in different scenarios could be poor due to the domain gap caused by medical equipment, shooting angle, etc. This paper proposed a scheme for model training enhancement through intermediate domains. Firstly, a residual neural network was utilized for a multi-classification task for fundus multi-disease detection. Secondly, CycleGAN was used as a classical-style migration generative network to generate target-like source domains. Finally, the intermediate domain was used to augment the source domain to train the deep residual network, thus implementing a UDA scheme. Experimental validation on two different domain datasets, On-site and Off-site, showed improved test results after augmentation with the intermediate domain. On the On-site dataset, the F1 Score and AUC improved by 2.72% and 2.94%, respectively. On the Off-site dataset, the F1 score and AUC improved by 4.32% and 6.78%, respectively. The experiments showed that the proposed scheme in this paper could effectively reduce the domain gap's impact and improve the model's generalization ability.

With the rapid development of biomedical and intelligent systems, more and more surgical modalities can be created and improved with computers. The finite element method is widely used in orthopaedics and can also simulate biomechanical changes in perianal soft tissues. This study created surgical models of perianal abscesses with different working conditions in SUPER-SAP finite element software by giving different tissue mechanical properties through MR image mapping and axial pulling tests. The results suggest that perianal abscess surgery with an incisional indwelling skin bridge may reduce displacement deformation at key anal nodes, revealing that the skin bridge may be one of the potential factors in preserving the functional reconstruction of the anus. This study may provide a basis for further exploration of perianal biomechanical mechanisms.

The ongoing COVID-19 pandemic continues to pose a grave danger to both human lives and the broader functioning of society. In response, healthcare organizations have increasingly turned to the powerful tools of artificial intelligence (AI) to tackle the complex challenges posed by the virus. In light of the overwhelming burden placed on the healthcare system during the COVID-19 pandemic, researchers have developed a broad range of AI algorithms and models to aid in clinical diagnosis, epidemic prediction, and clinical treatment. This study provides a comprehensive overview of the latest advancements in AI research in COVID-19, covering four key areas: auxiliary diagnosis and screening, auxiliary clinical therapy, prediction of disease severity and patient prognosis, and epidemic trend prediction. The ultimate aim of this study is to serve as a valuable resource for managing epidemics and informing effective prevention and treatment strategies for future outbreaks.

Decorrelation-based OCTA is a widely used optical coherence tomography angiography method which utilizes OCT intensity information. However, cardiac and respiratory motions in animals seriously degrade image quality. This kind of bulk motion is periodic, and its C-Scan (slow scan) direction component hinders motion correction because of scan position and OCT structure’s change. Some correction methods were proposed, but vasculature information will be lost when larger bulk motion occur. Here we demonstrate a correction method which uses stitch scan protocol in C-Scan direction, and sets a threshold to the maximum value of normalized cross-correlation among repeated B-Scan intensity signal to exclude false OCT B-scans. Result of In vivo imaging experiment for mice indicates that our method can reserve whole vasculature information and effectively improve image quality.

AS is one of the most common heart valve diseases, and early application of aortic valve replacement is currently the only way to treat AS. Preoperative evaluation and measurement of AS using one-stop CTA can provide clinicians with more valuable information. Currently, the sites measured by one-stop CTA include AA, STJ, AV, LCA height/RCA height, LVOT.

Currently, intelligent algorithms are widely used in the field of image processing to reduce the impact of fog on images. Surgical imaging systems are medical imaging devices used in surgical operations. With the continuous improvement of medical imaging technology and the application of advanced intelligent algorithms in medical devices, laparoscopic surgery is increasingly replacing traditional open surgery. However, due to the difference in temperature inside and outside the abdomen, lens fogging during laparoscopic surgery can affect the surgical field of view and interfere with the surgical process. The intelligent temperature balance control system, as a new technology, aims to improve the imaging field of view at different temperatures by combining hardware design with intelligent imaging algorithms. However, there is currently no evidence in the literature to support the anti-fog effect of surgical imaging systems equipped with this function. Therefore, this study aims to evaluate the advantages of surgical imaging systems equipped with the intelligent temperature balance control algorithm in clinical applications through comparative testing experiments and the use of Health Technology Assessment (HTA) tools and methods.