Photon-counting detectors are greatly improving the resolution and image quality in computed tomography (CT) technology. The drawback is, however, that the reconstruction becomes more challenging. This is because there is a considerable increment of the processing data due to the multiple energy bins and materials in the reconstruction analysis, as well as improved resolution. Yet efficient material decomposition and reconstruction methods tend to generate noisy images that do not completely satisfy the expected image quality. Therefore, there is a need for efficient denoising of the resulting material images. We present a new and fast denoiser that is based on a linear minimum mean square error (LMMSE) estimator. The LMMSE is very fast to compute, but not commonly used for CT image denoising, probably due to its inability to adapt the amount of denoising to different parts of the image and the difficulty to derive accurate statistical properties from the CT data. To overcome these problems we propose a model-based deep learning strategy, that is, a deep neural network that preserves an LMMSE structure (model-based), providing more robustness unseen data, as well as good interpretability to the result. In this way, the solution adapts to the anatomy in every point of the image and noise properties at that particular location. In order to asses the performance of the new method, we compare it to both to a conventional LMMSE estimator and to a “black-box” CNN in a simulation study with anthropomorphic phantoms.
Photon-counting spectral CT is a novel technology with a lot of promise. However, one common issue is detector inhomogeneity which results in streak artifacts in the sinogram domain and ring artifacts in the image domain. These rings are very conspicuous and limit the clinical usefulness of the images. We propose a deep learning based image processing technique for ring artifact correction in the sinogram domain. In particular, we train a UNet using a perceptual loss function with VGG16 as feature extractor to remove streak artifacts in the basis sinograms. Our results show that this method can successfully produce ring-corrected virtual monoenergetic images at a range of energy levels.
Photon-counting detectors (PCD) are the most recent advancement in computed tomography (CT). PCDs allow, among other things, for material decomposition, which decomposes the imaged object into a set of basis materials. Another field that is gaining attention, is the use of deep learning to improve the image reconstruction process in CT. In this work, we study the use of deep learning, specifically convolutional neural networks trained on the KiTS19 Challenge kidney data set, to improve the image quality of basis images resulting from three-material decomposition, a problem that is difficult due to its high sensitivity to noise. Our objective is to compare different network architectures and investigate whether these are best implemented in the projection domain or in the image domain. We study three different network architectures: U-Net, Dilated U-Net and ResNet, each applied in either the image domain or in the projection domain. The resulting image quality is evaluated in terms of contrast-to-noise ratio, task transfer function and noise power spectrum. Results show that for the type of phantoms the networks were trained on, the most effective option is to implement the network in the image domain and to use either the U-Net or Dilated U-Net architectures. However, when applying the networks to other phantoms, it seems that the networks in the sinogram generalize better, and produce better results. We also discuss why this might be the case, compare it with previous research, and consider what further improvements can be made.
Photon counting detectors in x-ray computed tomography (CT) are a major technological advancement that provides additional energy information and improves the decomposition of the CT image into material images. One important challenge in this new modality is how best to perform tomographic reconstruction. As a result of measuring multiple projections from different energy bins, more complex reconstruction algorithms are required. These are computationally demanding, due to the their large number of degrees of freedom. Also, the reconstruction algorithm needs to output multi-material image solutions. Reconstruction algorithms for spectral CT divide into two paradigms: two-step and one-step. Most typical solution is the two-step approach, where a first step consists of a material decomposition in projection domain, and a second step on tomographic reconstruction of each projection. This solution is computationally tractable but can cause a loss of information and it is difficult to regularize. The one-step solution, on the other hand, solves a joint optimization material decomposition and reconstruction, it is solved iteratively and can be, however, very time consuming. We present a deep learning-based solution to the one-step problem, with an architecture that mimics the updates of a primal-dual solver, and has demonstrated much greater computational efficiency than model-based iterative reconstruction. We have studied a proof-of-concept on a set of 700 Shepp-Logan phantoms. Our approach has shown enhanced performance compared to a model-based two-step approach, as well as compared to considering deep learning only in the first step of the two-step solution.
Photon counting detectors in x-ray computed tomography (CT) improve the decomposition of the CT scans into different materials. This decomposition is however not straightforward to solve, both in terms of computation expense and Photon counting detectors in x-ray computed tomography (CT) are a major technological advancement that provides additional energy information, and improve the decomposition of the CT image into material images. This material decomposition problem is however a non-linear inverse problem that is difficult to solve, both in terms of computation expense and accuracy. The most accepted solution consists in defining an optimization problem based on a maximum likelihood (ML) estimate with Poisson statistics, which is a model-based approach very dependent on the considered forward model and the chosen optimization solver. This may make the material decomposition result noisy and slow to be computed. To incorporate data-driven enhancement to the ML estimate, we propose a deep learning post-processing technique. Our approach is based on convolutional residual blocks that mimic the updates of an iterative optimization process and consider the ML estimate as an input. Therefore, our architecture implicitly considers the physical models of the problem, and in consequence needs less training data and fewer parameters than other standard convolutional networks typically used in medical imaging. We have studied a simulation case of our deep learning post-processing, first on a set of 350 Shepp-Logan -based phantoms, and then in a 600 human numerical phantoms. Our approach has shown denoising enhancement over two different ray-wise decomposition methods: one based on a Newton’s method to solve the ML estimation, and one based on a linear least-squares approximation of the ML expression. We believe this new deep learning post-processing approach is a promising technique to denoise material-decomposed sinograms in photon-counting CT.
Optical Coherence Tomography is a natural candidate for imaging biological structures just under tissue surface. Human thoracic aorta from aneurysms reveal elastin disorders and smooth muscle cell alterations when visualizing the media layer of the aortic wall, which is only some tens of microns in depth from surface. The resulting images require a suitable processing to enhance interesting disorder features and to use them as indicators for wall degradation, converting OCT into a hallmark for diagnosis of risk of aneurysm under intraoperative conditions. This work proposes gradient-based digital image processing approaches to conclude this risk. These techniques are believed to be useful in these applications as aortic wall disorders directly affect the refractive index of the tissue, having an effect on the gradient of the tissue reflectivity that conform the OCT image. Preliminary results show that the direction of the gradient contains information to estimate the tissue abnormality score. The detection of the edges of the OCT image is performed using the Canny algorithm. The edges delineate tissue disorders in the region of interest and isolate the abnormalities. These edges can be quantified to estimate a degradation score. Furthermore, the direction of the gradient seems to be a promising enhancement technique, as it detects areas of homogeneity in the region of interest. Automatic results from gradient-based strategies are finally compared to the histopathological global aortic score, which accounts for each risk factor presence and seriousness.
Optical coherence tomography images of human thoracic aorta from aneurysms reveal elastin disorders and smooth muscle cell alterations when visualizing the media layer of the aortic wall. These disorders can be employed as indicators for wall degradation and, therefore, become a hallmark for diagnosis of risk of aneurysm under intraoperative conditions. Two approaches are followed to evaluate this risk: the analysis of the reflectivity decay along the penetration depth and the textural analysis of a two-dimensional spatial distribution of the aortic wall backscattering. Both techniques require preprocessing stages for the identification of the air–sample interface and for the segmentation of the media layer. Results show that the alterations in the media layer of the aortic wall are better highlighted when the textural approach is considered and also agree with a semiquantitative histopathological grading that assesses the degree of wall degradation. The correlation of the co-occurrence matrix attains a sensitivity of 0.906 and specificity of 0.864 when aneurysm automatic diagnosis is evaluated with a receiver operating characteristic curve.
Optical coherence tomographic images of ascending thoracic human aortas from aneurysms exhibit disorders on the smooth muscle cell structure of the media layer of the aortic vessel as well as elastin degradation. Ex-vivo measurements of human samples provide results that correlate with pathologist diagnosis in aneurysmatic and control aortas. The observed disorders are studied as possible hallmarks for aneurysm diagnosis. To this end, the backscattering profile along the vessel thickness has been evaluated by fitting its decay against two different models, a third order polynomial fitting and an exponential fitting. The discontinuities present on the vessel wall on aneurysmatic aortas are slightly better identified with the exponential approach. Aneurysmatic aortic walls present uneven reflectivity decay when compared with healthy vessels. The fitting error has revealed as the most favorable indicator for aneurysm diagnosis as it provides a measure of how uniform is the decay along the vessel thickness.
Fractal analysis combined with a label-free scattering technique is proposed for describing the pathological architecture
of tumors. Clinicians and pathologists are conventionally trained to classify abnormal features such as structural
irregularities or high indices of mitosis. The potential of fractal analysis lies in the fact of being a morphometric measure
of the irregular structures providing a measure of the object’s complexity and self-similarity. As cancer is characterized
by disorder and irregularity in tissues, this measure could be related to tumor growth. Fractal analysis has been probed in
the understanding of the tumor vasculature network. This work addresses the feasibility of applying fractal analysis to
the scattering power map (as a physical modeling) and principal components (as a statistical modeling) provided by a
localized reflectance spectroscopic system. Disorder, irregularity and cell size variation in tissue samples is translated
into the scattering power and principal components magnitude and its fractal dimension is correlated with the pathologist
assessment of the samples. The fractal dimension is computed applying the box-counting technique. Results show that
fractal analysis of ex-vivo fresh tissue samples exhibits separated ranges of fractal dimension that could help classifier
combining the fractal results with other morphological features. This contrast trend would help in the discrimination of
tissues in the intraoperative context and may serve as a useful adjunct to surgeons.
Texture analysis of light scattering in tissue is proposed to obtain diagnostic information from breast cancer specimens. Light scattering measurements are minimally invasive, and allow the estimation of tissue morphology to guide the surgeon in resection surgeries. The usability of scatter signatures acquired with a micro-sampling reflectance spectral imaging system was improved utilizing an empirical approximation to the Mie theory to estimate the scattering power on a per-pixel basis. Co-occurrence analysis is then applied to the scattering power images to extract the textural features. A statistical analysis of the features demonstrated the suitability of the autocorrelation for the classification of notmalignant (normal epithelia and stroma, benign epithelia and stroma, inflammation), malignant (DCIS, IDC, ILC) and adipose tissue, since it reveals morphological information of tissue. Non-malignant tissue shows higher autocorrelation values while adipose tissue presents a very low autocorrelation on its scatter texture, being malignant the middle ground. Consequently, a fast linear classifier based on the consideration of just one straightforward feature is enough for providing relevant diagnostic information. A leave-one-out validation of the linear classifier on 29 samples with 48 regions of interest showed classification accuracies of 98.74% on adipose tissue, 82.67% on non-malignant tissue and 72.37% on malignant tissue, in comparison with the biopsy H and E gold standard. This demonstrates that autocorrelation analysis of scatter signatures is a very computationally efficient and automated approach to provide pathological information in real-time to guide surgeon during tissue resection.
A blind separation technique based on Independent Component Analysis (ICA) is proposed for breast tumor delineation
and pathologic diagnosis. Tissue morphology is determined by fitting local measures of tissue reflectance to a Mie
theory approximation, parameterizing the scattering power, scattering amplitude and average scattering irradiance. ICA
is applied on the scattering parameters by spatial analysis using the Fast ICA method to extract more determinant
features for an accurate diagnostic. Neither training, nor comparisons with reference parameters are required. Tissue
diagnosis is provided directly following ICA application to the scattering parameter images. Surgically resected breast
tissues were imaged and identified by a pathologist. Three different tissue pathologies were identified in 29 samples and
classified as not-malignant, malignant and adipose. Scatter plot analysis of both ICA results and optical parameters
where obtained. ICA subtle ameliorates those cases where optical parameter's scatter plots were not linearly separable.
Furthermore, observing the mixing matrix of the ICA, it can be decided when the optical parameters themselves are
diagnostically powerful. Moreover, contrast maps provided by ICA correlate with the pathologic diagnosis. The time
response of the diagnostic strategy is therefore enhanced comparing with complex classifiers, enabling near real-time
assessment of pathology during breast-conserving surgery.
A spectral analysis technique to enhance tumor contrast during breast conserving surgery is proposed. A set of 29
surgically-excised breast tissues have been imaged in local reflectance geometry. Measures of broadband reflectance are
directly analyzed using Principle Component Analysis (PCA), on a per sample basis, to extract areas of maximal spectral
variation. A dynamic selection threshold has been applied to obtain the final number of principal components,
accounting for inter-patient variability. A blind separation technique based on Independent Component Analysis (ICA) is
then applied to extract diagnostically powerful results. ICA application reveals that the behavior of one independent
component highly correlates with the pathologic diagnosis and it surpasses the contrast obtained using empirical models.
Moreover, blind detection characteristics (no training, no comparisons with training reference data) and no need for
parameterization makes the automated diagnosis simple and time efficient, favoring its translation to the clinical
practice. Correlation coefficient with model-based results up to 0.91 has been achieved.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.