This PDF file contains the front matter associated with SPIE Proceedings Volume 9420, including the Title Page, Copyright information, Table of Contents, Introduction (if any), and Conference Committee listing.

Characteristics of microvasculature can be revealed by immunohistochemical tissue staining, but manual quantification of these characteristics on whole-slide images containing potentially hundreds of vessels on each is tedious and subject to operator variability. Our objective was to develop and validate a fully automated segmentation of the vascular smooth muscle layer on whole-section histology of wild type and regenerated post-ischemia mouse hind limb microvasculature, stained for smooth muscle using 3,3'-Diaminobenzidine (DAB) immunostain. Major challenges to this segmentation are the irregularity of vessel wall staining, resulting in apparent fragmentation of intact vessels on the images, and artefactual appearance of stain on structures other than vessel walls. The automated segmentation localizes these fragments by color deconvolution to isolate the DAB stain. Complete vessel walls were reconstituted by joining of the topological skeletons of the vessel wall fragments. Based on a highly accurate registration of serial histology sections previously developed in our lab, artefactual fragments were removed based on measured incoherence with neighboring tissue in 3D. The vessel wall thickness, and vessel count, density, area, and perimeter were quantified. For segmentation validation, vessels were manually delineated and compared to the automated segmentation approach on a wild type mouse, with a dice similarity coefficient of 0.84. Differences were observed in the vessel measurements between the wild type and the regenerated vasculature in post-ischemic mice, as expected. Fully automatic and accurate measures of the vascular morphology are feasible with the automated segmentation of the vascular smooth muscle.

Immunohistochemistry (IHC) staining is an important technique for the detection of one or more biomarkers within a single tissue section. In digital pathology applications, the correct unmixing of the tissue image into its individual constituent dyes for each biomarker is a prerequisite for accurate detection and identification of the underlying cellular structures. A popular technique thus far is the color deconvolution method¹ proposed by Ruifrok et al. However, Ruifrok's method independently estimates the individual dye contributions at each pixel which potentially leads to “holes and cracks” in the cells in the unmixed images. This is clearly inadequate since strong spatial dependencies exist in the tissue images which contain rich cellular structures. In this paper, we formulate the unmixing algorithm into a least-square framework of image patches, and propose a novel color deconvolution method which explicitly incorporates the spatial smoothness and structure continuity constraint into a neighborhood graph regularizer. An analytical closed-form solution to the cost function is derived for this algorithm for fast implementation. The algorithm is evaluated on a clinical data set containing a number of 3,3-Diaminobenzidine (DAB) and hematoxylin (HTX) stained IHC slides and demonstrates better unmixing results than the existing strategy.

One of the main factors for high workload in pulmonary pathology in developing countries is the relatively large proportion of tuberculosis (TB) cases which can be detected with high throughput using automated approaches. TB is caused by Mycobacterium tuberculosis, which appears as thin, rod-shaped acid-fast bacillus (AFB) in Ziehl-Neelsen (ZN) stained sputum smear samples. In this paper, we present an algorithm for automatic detection of AFB in digitized images of ZN stained sputum smear samples under a light microscope. A key component of the proposed algorithm is the enhancement of raw input image using a novel anisotropic tubular filter (ATF) which suppresses the background noise while simultaneously enhancing strong anisotropic features of AFBs present in the image. The resulting image is then segmented using color features and candidate AFBs are identified. Finally, a support vector machine classifier using morphological features from candidate AFBs decides whether a given image is AFB positive or not. We demonstrate the effectiveness of the proposed ATF method with two different feature sets by showing that the proposed image analysis pipeline results in higher accuracy and F1-score than the same pipeline with standard median filtering for image enhancement.

It is our conjecture that the variability of colors in a pathology image effects the interpretation of pathology cases, whether it is diagnostic accuracy, diagnostic confidence, or workflow efficiency. In this paper, digital pathology images are analyzed to quantify the perceived difference in color that occurs due to display aging, in particular a change in the maximum luminance, white point, and color gamut. The digital pathology images studied include diagnostically important features, such as the conspicuity of nuclei. Three different display aging models are applied to images: aging of luminance and chrominance, aging of chrominance only, and a stabilized luminance and chrominance (i.e., no aging). These display models and images are then used to compare conspicuity of nuclei using CIE ΔE₂₀₀₀, a perceptual color difference metric. The effect of display aging using these display models and images is further analyzed through a human reader study designed to quantify the effects from a clinical perspective. Results from our reader study indicate significant impact of aged displays on workflow as well as diagnosis. Comparing original (not aged) images to aged images, aged images were significantly more difficult to read (p-value of 0.0005) and took longer to score (p-value of 0.02). Moreover, luminance and chrominance aging significantly reduced inter-session percent agreement of diagnostic scores (p-value of 0.0418).

We propose a method to detect and segment the oligodendrocytes and gliomas in OLIG2 immunoperoxidase stained tissue sections. Segmentation of cell nuclei is essential for automatic, fast, accurate and consistent analysis of pathology images. In general, glioma cells and oligodendrocytes mostly differ in shape and size within the tissue slide. In OLIG2 stained tissue images, gliomas are represented with irregularly shaped nuclei with varying sizes and brown shades. On the other hand, oligodendrocytes have more regular round nuclei shapes and are smaller in size when compared to glioma cells found in oligodendroglioma, astrocytomas, or oligoastrocytomas. The first task is to detect the OLIG2 positive cell regions within a region of interest image selected from a whole slide. The second task is to segment each cell nucleus and count the number of cell nuclei. However, the cell nuclei belonging to glioma cases have particularly irregular nuclei shapes and form cell clusters by touching or overlapping with each other. In addition to this clustered structure, the shading of the brown stain and the texture of the nuclei differ slightly within a tissue image. The final step of the algorithm is to classify glioma cells versus oligodendrocytes. Our method starts with color segmentation to detect positively stained cells followed by the classification of single individual cells and cell clusters by K-means clustering. Detected cell clusters are segmented with the H-minima based watershed algorithm. The novel aspects of our work are: 1) the detection and segmentation of multiple-type, positively-stained nuclei by incorporating only minimal prior information; and 2) adaptively determining clustering parameters to adjust to the natural variation in staining as well as the underlying cellular structure while accommodating multiple cell types in the image. Performance of the algorithm to detect individual cells is evaluated by sensitivity and precision metrics. Promising segmentation results (91% sensitivity and 86% precision) were achieved for a dataset of fourteen tissue slides with ground truth markings by two pathologists.

Neuroblastoma is a childhood cancer that starts in very early forms of nerve cells found in an embryo or fetus. It is a highly lethal cancer of sympathetic nervous system that commonly affects children of age five or younger. It accounts for a disproportionate number of childhood cancer deaths and remains a difficult cancer to eradicate despite intensive treatment that includes chemotherapy, surgery, hematopoietic stem cell transplantation, radiation therapy and immunotherapy. A poorly characterized group of patients are the 15% with primary refractory neuroblastoma (PRN) which is uniformly lethal due to de novo chemotherapy resistance. The lack of response to therapy is currently assessed after multiple months of cytotoxic therapy, driving the critical need to develop pretreatment clinic-biological biomarkers that can guide precise and effective therapeutic strategies. Therefore, our guiding hypothesis is that PRN has distinct biological features present at diagnosis that can be identified for prediction modeling. During a visual analysis of PRN slides, stained with hematoxylin and eosin, we observed that patients who survived for less than three years contained large eosin-stained structures as compared to those who survived for greater than three years. So, our hypothesis is that the size of eosin stained structures can be used as a differentiating feature to characterize recurrence in neuroblastoma. To test this hypothesis, we developed an image analysis method that performs stain separation, followed by the detection of large structures stained with Eosin. On a set of 21 PRN slides, stained with hematoxylin and eosin, our image analysis method predicted the outcome with 85.7% accuracy.

We examined and established the potential of ex-vivo confocal fluorescence microscopy for differentiating between normal cervical tissue, low grade Cervical Intraepithelial Neoplasia (CIN1), and high grade CIN (CIN2 and CIN3). Our objectives were to i) use Quantitative Tissue Phenotype (QTP) analysis to quantify nuclear and cellular morphology and tissue architecture in confocal microscopic images of fresh cervical biopsies and ii) determine the accuracy of high grade CIN detection via confocal microscopy. Cervical biopsy specimens of colposcopically normal and abnormal tissues obtained from 15 patients were evaluated by confocal fluorescence microscopy. Confocal images were analyzed and about 200 morphological and architectural features were calculated at the nuclear, cellular, and tissue level. For the purpose of this study, we used four features to delineate disease grade including nuclear size, cell density, estimated nuclear-cytoplasmic (ENC) ratio, and the average of three nearest Delaunay neighbors distance (3NDND). Our preliminary results showed ENC ratio and 3NDND correlated well with histopathological diagnosis. The Spearman correlation coefficient between each of these two features and the histopathological diagnosis was higher than the correlation coefficient between colposcopic appearance and histopathological diagnosis. Sensitivity and specificity of ENC ratio for detecting high grade CIN were both equal to 100%. QTP analysis of fluorescence confocal images shows the potential to discriminate high grade CIN from low grade CIN and normal tissues. This approach could be used to help clinicians identify HGSILs in clinical settings.

The number of new cases of thyroid cancer are dramatically increasing as incidences of this cancer have more than doubled since the early 1970s. Tall cell variant (TCV-PTC) papillary thyroid carcinoma is one type of thyroid cancer that is more aggressive and usually associated with higher local recurrence and distant metastasis. This variant can be identified through visual characteristics of cells in histological images. Thus, we created a fully automatic algorithm that is able to segment cells using a multi-stage approach. Our method learns the statistical characteristics of nuclei and cells during the segmentation process and utilizes this information for a more accurate result. Furthermore, we are able to analyze the detected regions and extract characteristic cell data that can be used to assist in clinical diagnosis.

Automated detection of prostate cancer in digitized H and E whole-slide images is an important first step for computer-driven grading. Most automated grading algorithms work on preselected image patches as they are too computationally expensive to calculate on the multi-gigapixel whole-slide images. An automated multi-resolution cancer detection system could reduce the computational workload for subsequent grading and quantification in two ways: by excluding areas of definitely normal tissue within a single specimen or by excluding entire specimens which do not contain any cancer. In this work we present a multi-resolution cancer detection algorithm geared towards the latter. The algorithm methodology is as follows: at a coarse resolution the system uses superpixels, color histograms and local binary patterns in combination with a random forest classifier to assess the likelihood of cancer. The five most suspicious superpixels are identified and at a higher resolution more computationally expensive graph and gland features are added to refine classification for these superpixels. Our methods were evaluated in a data set of 204 digitized whole-slide H and E stained images of MR-guided biopsy specimens from 163 patients. A pathologist exhaustively annotated the specimens for areas containing cancer. The performance of our system was evaluated using ten-fold cross-validation, stratified according to patient. Image-based receiver operating characteristic (ROC) analysis was subsequently performed where a specimen containing cancer was considered positive and specimens without cancer negative. We obtained an area under the ROC curve of 0.96 and a 0.4 specificity at a 1.0 sensitivity.

Studies have shown that different cell types of ovarian carcinoma have different molecular profiles, exhibit different behavior, and that patients could benefit from typespecific treatment. Different cell types display different histopathology features, and different criteria are used for each cell type classification. Inter-observer variability for the task of classifying ovarian cancer cell types is an under-examined area of research. This study served as a pilot study to quantify observer variability related to the classification of ovarian cancer cell types and to extract valuable data for designing a validation study of digital pathology (DP) for this task. Three observers with expertise in gynecologic pathology reviewed 114 cases of ovarian cancer with optical microscopy, with specific guidelines for classifications into distinct cell types. For 93 cases all 3 pathologists agreed on the same cell type, for 18 cases 2 out of 3 agreed, and for 3 cases there was no agreement. Across cell types with a minimum sample size of 10 cases, agreement between all three observers was {91.1%, 80.0%, 90.0%, 78.6%, 100.0%, 61.5%} for the high grade serous carcinoma, low grade serous carcinoma, endometrioid, mucinous, clear cell, and carcinosarcoma cell types respectively. These results indicate that unanimous agreement varied over a fairly wide range. However, additional research is needed to determine the importance of these differences in comparison studies. These results will be used to aid in the design and sizing of such a study comparing optical and digital pathology. In addition, the results will help in understanding the potential role computer-aided diagnosis has in helping to improve the agreement of pathologists for this task.

Immunofluorescent (IF) image analysis of tissue pathology has proven to be extremely valuable and robust in developing prognostic assessments of disease, particularly in prostate cancer. There have been significant advances in the literature in quantitative biomarker expression as well as characterization of glandular architectures in discrete gland rings. However, while biomarker and glandular morphometric features have been combined as separate predictors in multivariate models, there is a lack of integrative features for biomarkers co-localized within specific morphological sub-types; for example the evaluation of androgen receptor (AR) expression within Gleason 3 glands only. In this work we propose a novel framework employing multiple techniques to generate integrated metrics of morphology and biomarker expression. We demonstrate the utility of the approaches in predicting clinical disease progression in images from 326 prostate biopsies and 373 prostatectomies. Our proposed integrative approaches yield significant improvements over existing IF image feature metrics. This work presents some of the first algorithms for generating innovative characteristics in tissue diagnostics that integrate co-localized morphometry and protein biomarker expression.

Visual estimation of tumor and stroma proportions in microscopy images yields a strong, Tumor-(lymph)Node- Metastasis (TNM) classification-independent predictor for patient survival in colorectal cancer. Therefore, it is also a potent (contra)indicator for adjuvant chemotherapy. However, quantification of tumor and stroma through visual estimation is highly subject to intra- and inter-observer variability. The aim of this study is to develop and clinically validate a method for objective quantification of tumor and stroma in standard hematoxylin and eosin (H and E) stained microscopy slides of rectal carcinomas. A tissue segmentation algorithm, based on supervised machine learning and pixel classification, was developed, trained and validated using histological slides that were prepared from surgically excised rectal carcinomas in patients who had not received neoadjuvant chemotherapy and/or radiotherapy. Whole-slide scanning was performed at 20× magnification. A total of 40 images (4 million pixels each) were extracted from 20 whole-slide images at sites showing various relative proportions of tumor and stroma. Experienced pathologists provided detailed annotations for every extracted image. The performance of the algorithm was evaluated using cross-validation by testing on 1 image at a time while using the other 39 images for training. The total classification error of the algorithm was 9.4% (SD = 3.2%). Compared to visual estimation by pathologists, the algorithm was 7.3 times (P = 0.033) more accurate in quantifying tissues, also showing 60% less variability. Automatic tissue quantification was shown to be both reliable and practicable. We ultimately intend to facilitate refined prognostic stratification of (colo)rectal cancer patients and enable better personalized treatment.

In the popular Nottingham histologic score system for breast cancer grading, the pathologist analyzes the H and E tissue slides and assigns a score, in the range of 1-3, for tubule formation, nuclear pleomorphism and mitotic activity in the tumor regions. The scores from these three factors are added to give a final score, ranging from 3-9 to grade the cancer. Tubule score (TS), which reflects tubular formation, is a value in 1-3 given by manually estimating the percentage of glandular regions in the tumor that form tubules. In this paper, given an H and E tissue image representing a tumor region, we propose an automated algorithm to detect glandular regions and detect the presence of tubules in these regions. The algorithm first detects all nuclei and lumen candidates in the input image, followed by identifying tumor nuclei from the detected nuclei and identifying true lumina from the lumen candidates using a random forest classifier. Finally, it forms the glandular regions by grouping the closely located tumor nuclei and lumina using a graph-cut-based method. The glandular regions containing true lumina are considered as the ones that form tubules (tubule regions). To evaluate the proposed method, we calculate the tubule percentage (TP), i.e., the ratio of the tubule area to the total glandular area for 353 H and E images of the three TSs, and plot the distribution of these TP values. This plot shows the clear separation among these three scores, suggesting that the proposed algorithm is useful in distinguishing images of these TSs.

This paper presents a new algorithm for automatic detection of regions of interest in whole slide histopathological images. The proposed algorithm generates and classifies superpixels at multiple resolutions to detect regions of interest. The algorithm emulates the way the pathologist examines the whole slide histopathology image by processing the image at low magnifications and performing more sophisticated analysis only on areas requiring more detailed information. However, instead of the traditional usage of fixed sized rectangular patches for the identification of relevant areas, we use superpixels as the visual primitives to detect regions of interest. Rectangular patches can span multiple distinct structures, thus degrade the classification performance. The proposed multi-scale superpixel classification approach yields superior performance for the identification of the regions of interest. For the evaluation, a set of 10 whole slide histopathology images of breast tissue were used. Empirical evaluation of the performance of our proposed algorithm relative to expert manual annotations shows that the algorithm achieves an area under the Receiver operating characteristic (ROC) curve of 0.958, demonstrating its efficacy for the detection of regions of interest.

Breast cancer diagnosis relies on staining serial sections of a biopsy in a process that can be time intensive and costly. Fourier transform infrared imaging (FT-IR) is a non-destructive, label-free chemical imaging technique that uses the vibrational structure of the biological molecules of the sample to provide contrast for images at any absorption peak in the mid-infrared. The full potential of spectroscopic imaging has been limited by the spatial resolution provided by most commercial instruments. By increasing the magnification and numerical aperture of the microscope, image pixel sizes on the order of 1.1 micron can be achieved, allowing HD FT-IR spectroscopic imaging to provide high quality images that could aid in histopathology, diagnosis, and studies of breast cancer progression.

Stain separation is the process whereby a full colour histology section image is transformed into a series of single channel images, each corresponding to a given stain's expression. Many algorithms in the field of digital pathology are concerned with the expression of a single stain, thus stain separation is a key preprocessing step in these situations. We present a new versatile method of stain separation. The method uses Independent Component Analysis (ICA) to determine a set of statistically independent vectors, corresponding to the individual stain expressions. In comparison to other popular approaches, such as PCA and NNMF, we found that ICA gives a superior projection of the data with respect to each stain. In addition, we introduce a correction step to improve the initial results provided by the ICA coefficients. Many existing approaches only consider separation of two stains, with primary emphasis on Haematoxylin and Eosin. We show that our method is capable of making a good separation when there are more than two stains present. We also demonstrate our method's ability to achieve good separation on a variety of different stain types.

There is currently an increasing interest in combining the information obtained from radiology and histology with the intent of gaining a better understanding of how different tumour morphologies can lead to distinctive radiological signs which might predict overall treatment outcome. Relating information at different resolution scales is challenging. Reconstructing 3D volumes from histology images could be the key to interpreting and relating the radiological image signal to tissue microstructure. The goal of this study is to determine the minimum sampling (maximum spacing between histological sections through a fixed surgical specimen) required to create a 3D reconstruction of the specimen to a specific tolerance. We present initial results for one lumpectomy specimen case where 33 consecutive histology slides were acquired.

Digital cancer diagnosis is a research realm where signal processing techniques are used to analyze and to classify color histopathology images. Different from grayscale image analysis of magnetic resonance imaging or X-ray, colors in histopathology images convey large amount of histological information and thus play significant role in cancer diagnosis. Though color information is widely used in histopathology works, as today, there is few study on color model selections for feature extraction in cancer diagnosis schemes. This paper addresses the problem of color space selection for digital cancer classification using H and E stained images, and investigates the effectiveness of various color models (RGB, HSV, CIE L*a*b*, and stain-dependent H and E decomposition model) in breast cancer diagnosis. Particularly, we build a diagnosis framework as a comparison benchmark and take specific concerns of medical decision systems into account in evaluation. The evaluation methodologies include feature discriminate power evaluation and final diagnosis performance comparison. Experimentation on a publicly accessible histopathology image set suggests that the H and E decomposition model outperforms other assessed color spaces. For reasons behind various performance of color spaces, our analysis via mutual information estimation demonstrates that color components in the H and E model are less dependent, and thus most feature discriminate power is collected in one channel instead of spreading out among channels in other color spaces.

Reliable automated prostate tumor detection and characterization in whole-mount histology images is sought in many applications, including post-resection tumor staging and as ground-truth data for multi-parametric MRI interpretation. In this study, an ensemble-based supervised classification algorithm for high-resolution histology images was trained on tile-based image features including histogram and gray-level co-occurrence statistics. The algorithm was assessed using different combinations of H and E prostate slides from two separate medical centers and at two different magnifications (400x and 200x), with the aim of applying tumor classification models to new data. Slides from both datasets were annotated by expert pathologists in order to identify homogeneous cancerous and non-cancerous tissue regions of interest, which were then categorized as (1) low-grade tumor (LG-PCa), including Gleason 3 and high-grade prostatic intraepithelial neoplasia (HG-PIN), (2) high-grade tumor (HG-PCa), including various Gleason 4 and 5 patterns, or (3) non-cancerous, including benign stroma and benign prostatic hyperplasia (BPH). Classification models for both LG-PCa and HG-PCa were separately trained using a support vector machine (SVM) approach, and per-tile tumor prediction maps were generated from the resulting ensembles. Results showed high sensitivity for predicting HG-PCa with an AUC up to 0.822 using training data from both medical centres, while LG-PCa showed a lower sensitivity of 0.763 with the same training data. Visual inspection of cancer probability heatmaps from 9 patients showed that 17/19 tumors were detected, and HG-PCa generally reported less false positives than LG-PCa.

Computerized histopathology image analysis enables an objective, efficient, and quantitative assessment of digitized histopathology images. Such analysis often requires an accurate and efficient detection and segmentation of histological structures such as glands, cells and nuclei. The segmentation is used to characterize tissue specimens and to determine the disease status or outcomes. The segmentation of nuclei, in particular, is challenging due to the overlapping or clumped nuclei. Here, we propose a nuclei seed detection method for the individual and overlapping nuclei that utilizes the gradient orientation or direction information. The initial nuclei segmentation is provided by a multiview boosting approach. The angle of the gradient orientation is computed and traced for the nuclear boundaries. Taking the first derivative of the angle of the gradient orientation, high concavity points (junctions) are discovered. False junctions are found and removed by adopting a greedy search scheme with the goodness-of-fit statistic in a linear least squares sense. Then, the junctions determine boundary segments. Partial boundary segments belonging to the same nucleus are identified and combined by examining the overlapping area between them. Using the final set of the boundary segments, we generate the list of seeds in tissue images. The method achieved an overall precision of 0.89 and a recall of 0.88 in comparison to the manual segmentation.

Computer-Aided Diagnosis (CAD) systems based on histopathological images rely on quality low-level image processing, including cell segmentation. Many methods for cell segmentation lack in generality and struggle with the wide variety of cell appearance and inter-cell structure present in histopathological images. We present a computationally efficient system to classify segmentation results as the first step toward automatic segment correction. This general method can applied to existing or future cell segmentation methods to provide corrections for low-quality results. Specifically, with a small collection of easy-to-compute features, we can identify incorrect segments with a high degree of accuracy, which then can be used to determine the needed corrections based on the type of segmentation failure present.

Automatic whole slide (WS) tissue image segmentation is an important problem in digital pathology. A conventional classification-based method (referred to as CCb method) to tackle this problem is to train a classifier on a pre-built training database (pre-built DB) obtained from a set of training WS images, and use it to classify all image pixels or image patches (test samples) in the test WS image into different tissue types. This method suffers from a major challenge in WS image analysis: the strong inter-slide tissue variability (ISTV), i.e., the variability of tissue appearance from slide to slide. Due to this ISTV, the test samples are usually very different from the training data, which is the source of misclassification. To address the ISTV, we propose a novel method, called slide-adapted classification (SAC), to extend the CCb method. We assume that in the test WS image, besides regions with high variation from the pre-built DB, there are regions with lower variation from this DB. Hence, the SAC method performs a two-stage classification: first classifies all test samples in a WS image (as done in the CCb method) and compute their classification confidence scores. Next, the samples classified with high confidence scores (samples being reliably classified due to their low variation from the pre-built DB) are combined with the pre-built DB to generate an adaptive training DB to reclassify the low confidence samples. The method is motivated by the large size of the test WS image (a large number of high confidence samples are obtained), and the lower variability between the low and high confidence samples (both belonging to the same WS image) compared to the ISTV. Using the proposed SAC method to segment a large dataset of 24 WS images, we improve the accuracy over the CCb method.

Pancreatic cancer is the fourth leading cause of cancer death in the United States. Most pancreatic cancer patients will die within the first year of diagnosis, and just 6% will survive five years. Currently, surgery is the only treatment that offers a chance of cure for pancreatic cancer patients. Accurately identifying the tumors margins in real time is a significant difficulty during pancreatic cancer surgery and contributes to the low 5-year survival rate. We are developing a hyperspectral imaging system based on compressive sampling for real-time tumor margin detection to facilitate more effective removal of diseased tissue and result in better patient outcomes. Recent research has shown that optical spectroscopy can be used to distinguish between healthy and diseased tissue and will likely become an important minimally invasive diagnostic tool for a range of diseases. Reflectance spectroscopy provides information about tissue morphology, while laser-induced autofluorescence spectra give accurate information about the content and molecular structure of the emitting tissue. We are developing a spectral imaging system that targets emission from collagen and NAD(P)H as diagnostics for differentiating healthy and diseased pancreatic tissue. In this study, we demonstrate the ability of our camera system to acquire hyperspectral images and its potential application for imaging autofluorescent emission from pancreatic tissue.

Along with mitotic count, nuclear pleomorphism or nuclear atypia is an important criterion for the grading of breast cancer in histopathology. Though some works have been done in mitosis detection (ICPR 2012,¹ MICCAI 2013,² and ICPR 2014), not much work has been dedicated to automated nuclear atypia grading, especially the most difficult task of detection of grade 3 nuclei. We propose the use of Convolutional Neural Networks for the automated detection of cell nuclei, using images from the three grades of breast cancer for training. The images were obtained from ICPR contests. Additional manual annotation was performed to classify pixels into five classes: stroma, nuclei, lymphocytes, mitosis and fat. At total of 3,000 thumbnail images of 101 × 101 pixels were used for training. By dividing this training set in an 80/20 ratio we could obtain good training results (around 90%). We tested our CNN on images of the three grades which were not in the training set. High grades nuclei were correctly classified. We then thresholded the classification map and performed basic analysis to keep only rounded objects. Our results show that mostly all atypical nuclei were correctly detected.

The first step prior to most analyses on most histopathology images is the detection of area of interest. In this work, we present a superpixel-based approach for glandular structure detection in colon histology images. An image is first segmented into superpixels with the constraint on the presence of glandular boundaries. Texture and color information is then extracted from each superpixel to calculate the probability of that superpixel belonging to glandular regions, resulting in a glandular probability map. In addition, we present a novel texture descriptor derived from a region covariance matrix of scattering coefficients. Our approach shows encouraging results for the detection of glandular structures in colon tissue samples.

The assessment of megakaryocytes (MKs) in bone marrow trephine images is an important step in the classification of different subtypes of myeloproliferative neoplasms (MPNs). In general, bone marrow trephine images include several types of cells mixed together, which make it quite difficult to visually identify MKs. In order to aid hematopathologists in the identification and study of MKs, we develop an image processing framework with supervised machine learning approaches and a novel circumscribing active contour model to identify potential MKs and then to accurately delineate the corresponding nucleus and membrane. Specifically, a number of color and texture features are used in a nave Bayesian classifier and an Adaboost classifier to locate the regions with a high probability of depicting MKs. A region-based active contour is used on the candidate MKs to accurately delineate the boundaries of nucleus and membrane. The proposed circumscribing active contour model employs external forces not only based on pixel intensities, but also on the probabilities of depicting MKs as computed by the classifiers. Experimental results suggest that the machine learning approach can detect potential MKs with an accuracy of more than 75%. When our circumscribing active contour model is employed on the candidate MKs, the nucleus and membrane boundaries are segmented with an accuracy of more than 80% as measured by the Dice similarity coefficient. Compared to traditional region-based active contours, the use of additional external forces based on the probability of depicting MKs improves segmentation performance and computational time by an average 5%.

Can we create values and make success by creating digitized Pathology by only changing the technical platform within the department of pathology? The answer is definitely No. To be able to create values that will pay for the investment of going digital we need to go outside traditional ways of change-management in healthcare. Cooperation before buying the hardware is the key, going there by creating a unique negotiated information-model that spans the whole value chain within the care process in regard of pathology services. This is the core of creating great values throughout the whole healthcare sub process of cancer detection. Region Västra Götaland (VGR), has taken this path and will show that it´s possible working/thinking outside the “box.”

We recently completed a reader study to compare optical and digital pathology (DP) for the assessment of two tissue-based biomarkers with immunohistochemistry. Eight pathologists reviewed 50 breast cancer whole slides (25 stained with HER2 and 25 with Ki-67) and 2 TMAs (1 stained with HER2, 1 with Ki-67, 97 cores each), using digital and optical microscopy. All reviews took place in a single office, using the same microscope, same computer/color calibrated monitor combination, and the same ambient light, in order to eliminate sources of variability due to these parameters. Agreement analysis was performed using the Kendall’s tau-b metric and percent correct agreement. Results showed relatively high overall inter-observer and inter-modality agreement. However, significant uncertainty was observed for the whole slide evaluation with 95% confidence intervals (CI) in the order of 0.30 for the Kendall’s tau-b metric, despite taking care to reduce sources of uncertainty. For the better-sampled TMAs, CIs were in the order of 0.15. It can be deduced that the sample size of 25 slides for each biomarker was not adequate even though it is in line with recent guidelines for the validation of DP from the College of American Pathologists (20 slides for immunohistochemistry without specifying task). Significant uncertainty was observed in our study, despite controlling for several variables. Further work is needed to identify sources of uncertainty for observer tasks in DP, and to account for it in study designs to assess DP.

Reliable segmentation of abnormal nuclei in cervical cytology is of paramount importance in automation-assisted screening techniques. This paper presents a general method for improving the segmentation of abnormal nuclei using a graph-search based approach. More specifically, the proposed method focuses on the improvement of coarse (initial) segmentation. The improvement relies on a transform that maps round-like border in the Cartesian coordinate system into lines in the polar coordinate system. The costs consisting of nucleus-specific edge and region information are assigned to the nodes. The globally optimal path in the constructed graph is then identified by dynamic programming. We have tested the proposed method on abnormal nuclei from two cervical cell image datasets, Herlev and H and E stained liquid-based cytology (HELBC), and the comparative experiments with recent state-of-the-art approaches demonstrate the superior performance of the proposed method.

The purpose of this study is to investigate the feasibility of applying automatic interphase FISH cells analysis method for detecting the residual malignancy of post chemotherapy leukemia patients. In the experiment, two clinical specimens with translocation between chromosome No. 9 and 22 or No. 11 and 14 were selected from the patients underwent leukemia diagnosis and treatment. The entire slide of each specimen was first digitalized by a commercial fluorescent microscope using a 40× objective lens. Then, the scanned images were processed by a computer-aided detecting (CAD) scheme to identify the analyzable FISH cells, which is accomplished by applying a series of features including the region size, Brenner gradient and maximum intensity. For each identified cell, the scheme detected and counted the number of the FISH signal dots inside the nucleus, using the adaptive threshold of the region size and distance of the labeled FISH dots. The results showed that the new CAD scheme detected 8093 and 6675 suspicious regions of interest (ROI) in two specimens, among which 4546 and 3807 ROI contain analyzable interphase FISH cell. In these analyzable ROIs, CAD selected 334 and 405 residual malignant cancer cells, which is substantially more than those visually detected in a cytogenetic laboratory of our medical center (334 vs. 122, 405 vs. 160). This investigation indicates that an automatic interphase FISH cell scanning and CAD method has the potential to improve the accuracy and efficiency of the prognostic assessment for leukemia and other genetic related cancer patients in the future.

The steatosis in liver pathological tissue images is a promising indicator of nonalcoholic fatty liver disease (NAFLD) and the possible risk of hepatocellular carcinoma (HCC). The resulting values are also important for ensuring the automatic and accurate classification of HCC images, because the existence of many fat droplets is likely to create errors in quantifying the morphological features used in the process. In this study we propose a method that can automatically detect, and exclude regions with many fat droplets by using the feature values of colors, shapes and the arrangement of cell nuclei. We implement the method and confirm that it can accurately detect fat droplets and quantify the fat droplet ratio of actual images. This investigation also clarifies the effective characteristics that contribute to accurate detection.

In Digital Pathology, one of the most simple and yet most useful feature is the ability to view serial sections of tissue simultaneously on a computer monitor. This enables the pathologist to evaluate the histology and expression of multiple markers for a patient in a single review. However, the rate limiting step in this process is the time taken for the pathologist to open each individual image, align the sections within the viewer, with a maximum of four slides at a time, and then manually move around the section. In addition, due to tissue processing and pre-analytical steps, sections with different stains have non-linear variations between the two acquisitions, that is, they will stretch and change shape from section to section. To date, no solution has come close to a workable solution to automatically align the serial sections into one composite image. This research work address this problem to obtain an automated serial section alignment tool enabling the pathologists to simply scroll through the various sections in a single viewer. To this aim a multi-resolution intensity-based registration method using mutual information as a similarity metric, an optimizer based on an evolutionary process and a bilinear transformation has been used. To characterize the performance of the algorithm 40 cases x 5 different serial sections stained with hematoxiline-eosine (HE), estrogen receptor (ER), progesterone receptor (PR), Ki67 and human epidermal growth factor receptor 2 (Her2), have been considered. The qualitative results obtained are promising, with average computation time of 26.4s for up to 14660x5799 images running interpreted code.

In this paper, we proposed a method based on the Freeman chain code to segment and count rhesus choroid-retinal vascular endothelial cells (RF/6A) automatically for fluorescence microscopy images. The proposed method consists of four main steps. First, a threshold filter and morphological transform were applied to reduce the noise. Second, the boundary information was used to generate the Freeman chain codes. Third, the concave points were found based on the relationship between the difference of the chain code and the curvature. Finally, cells segmentation and counting were completed based on the characteristics of the number of the concave points, the area and shape of the cells. The proposed method was tested on 100 fluorescence microscopic cell images, and the average true positive rate (TPR) is 98.13% and the average false positive rate (FPR) is 4.47%, respectively. The preliminary results showed the feasibility and efficiency of the proposed method.

In this work we present a time-lapsed confocal microscopy image analysis technique for an automated gene expression study of multiple single living cells. Fluorescence Resonance Energy Transfer (FRET) is a technology by which molecule-to-molecule interactions are visualized. We analyzed a dynamic series of ~10² images obtained using confocal microscopy of fluorescence in yeast cells containing RNA reporters that give a FRET signal when the gene promoter is activated. For each time frame, separate images are available for three spectral channels and the integrated intensity snapshot of the system. A large number of time-lapsed frames must be analyzed to identify each cell individually across time and space, as it is moving in and out of the focal plane of the microscope. This makes it a difficult image processing problem. We have proposed an algorithm here, based on scale-space technique, which solves the problem satisfactorily. The algorithm has multiple directions for even further improvement. The ability to rapidly measure changes in gene expression simultaneously in many cells in a population will open the opportunity for real-time studies of the heterogeneity of genetic response in a living cell population and the interactions between cells that occur in a mixed population, such as the ones found in the organs and tissues of multicellular organisms.