Although deep learning (DL) has demonstrated impressive diagnostic performance for a variety of computational pathology tasks, this performance often markedly deteriorates on whole slide images (WSI) generated at external test sites. This phenomenon is due in part to domain shift, wherein differences in test-site pre-analytical variables (e.g., slide scanner, staining procedure) result in WSI with notably different visual presentations compared to training data. To ameliorate pre-analytic variances, approaches such as CycleGAN can be used to calibrate visual properties of images between sites, with the intent of improving DL classifier generalizability. In this work, we present a new approach termed Multi-Site Cross-Organ Calibration based Deep Learning (MuSClD) that employs WSIs of an off-target organ for calibration created at the same site as the on-target organ, based off the assumption that cross-organ slides are subjected to a common set of pre-analytical sources of variance. We demonstrate that by using an off-target organ from the test site to calibrate training data, the domain shift between training and testing data can be mitigated. Importantly, this strategy uniquely guards against potential data leakage introduced during calibration, wherein information only available in the testing data is imparted on the training data. We evaluate MuSClD in the context of the automated diagnosis of non-melanoma skin cancer (NMSC). Specifically, we evaluated MuSClD for identifying and distinguishing (a) basal cell carcinoma (BCC), (b) in-situ squamous cell carcinomas (SCC-In Situ), and (c) invasive squamous cell carcinomas (SCC-Invasive), using an Australian (training, n = 85) and a Swiss (held-out testing, n = 352) cohort. Our experiments reveal that MuSCID reduces the Wasserstein distances between sites in terms of color, contrast, and brightness metrics, without imparting noticeable artifacts to training data. The NMSC-subtyping performance is statistically improved as a result of MuSCID in terms of one-vs. rest AUC: BCC (0.92 vs 0.87, p = 0.01), SCC-In Situ (0.87 vs 0.73, p = 0.15) and SCC-Invasive (0.92 vs 0.82, p = 1e-5). Compared to baseline NMSC-subtyping with no calibration, the internal validation results of MuSClD (BCC (0.98), SCC-In Situ (0.92), and SCC-Invasive (0.97)) suggest that while domain shift indeed degrades classification performance, our on-target calibration using off-target tissue can safely compensate for pre-analytical variabilities, while improving the robustness of the model.

Localized disease heterogeneity on imaging extracted via radiomics approaches have recently been associated with disease prognosis and treatment response. Traditionally, radiomics analyses leverage texture operators to derive voxel- or region-wise feature values towards quantifying subtle variations in image appearance within a region-of-interest (ROI). With the goal of mining additional voxel-wise texture patterns from radiomic "expression maps", we introduce a new RADIomic Spatial TexturAl descripTor (RADISTAT). This was driven by the hypothesis that quantifying spatial organization of texture patterns within an ROI could allow for better capturing interactions between different tissue classes present in a given region; thus enabling more accurate characterization of disease or response phenotypes. RADISTAT involves: (a) robustly identifying sub-compartments of low, intermediate, and high radiomic expression (i.e. heterogeneity) in a feature map and (b) quantifying spatial organization of sub-compartments via graph interactions. RADISTAT was evaluated in two clinically challenging problems: (1) discriminating nodal/distant metastasis from metastasis-free rectal cancer patients on post-chemoradiation T2w MRI, and (2) distinguishing tumor progression from pseudo-progression in glioblastoma multiforme using post-chemoradiation T1w MRI. Across over 800 experiments, RADISTAT yielded a consistent discriminatory signature for tumor progression (GBM) and disease metastasis (RCa); where its sub-compartments were associated with pathologic tissue types (fibrosis or tumor, determined via fusion of MRI and pathology). In a multi-institutional setting for both clinical problems, RADISTAT resulted in higher classifier performance (11% improvement in AUC, on average) compared to radiomic descriptors. Furthermore, combining RADISTAT with radiomic descriptors resulted in significantly improved performance compared to using radiomic descriptors alone.

The concept of tumor field effect implies that cancer is a systemic disease with its impact way beyond the visible tumor confines. For instance, in Glioblastoma (GBM), an aggressive brain tumor, the increase in intracranial pressure due to tumor burden often leads to brain herniation and poor outcomes. Our work is based on the rationale that highly aggressive tumors tend to grow uncontrollably, leading to pronounced biomechanical tissue deformations in the normal parenchyma, which when combined with local morphological differences in the tumor confines on MRI scans, will comprehensively capture tumor field effect. Specifically, we present an integrated MRI-based descriptor, radiomic-Deformation and Textural Heterogeneity (r-DepTH). This descriptor comprises measurements of the subtle perturbations in tissue deformations throughout the surrounding normal parenchyma due to mass effect. This involves non-rigidly aligning the patients' MRI scans to a healthy atlas via diffeomorphic registration. The resulting inverse mapping is used to obtain the deformation field magnitudes in the normal parenchyma. These measurements are then combined with a 3D texture descriptor, Co-occurrence of Local Anisotropic Gradient Orientations (COLLAGE), which captures the morphological heterogeneity and infiltration within the tumor confines, on MRI scans. In this work, we extensively evaluated r-DepTH for survival risk-stratification on a total of 207 GBM cases from 3 different cohorts (Cohort 1 ( n1 = 53 ), Cohort 2 ( n2 = 75 ), and Cohort 3 ( n3 = 79 )), where each of these three cohorts was used as a training set for our model separately, and the other two cohorts were used for testing, independently, for each training experiment. When employing Cohort 1 for training, r-DepTH yielded Concordance indices (C-indices) of 0.7 and 0.65, hazard ratios (HR) and Confidence Intervals (CI) of 10 (6 - 19) and 5 (3 - 8) on Cohorts 2 and 3, respectively. Similarly, training on Cohort 2 yielded C-indices of 0.6 and 0.7, HR and CI of 1 (0.7 - 2) and 3 (2 - 5) on Cohorts 1 and 3, respectively. Finally, training on Cohort 3 yielded C-indices of 0.75 and 0.63, HR and CI of 24 (10 - 57) and 12 (6 - 21) on Cohorts 1 and 2, respectively. Our results show that r-DepTH descriptor may serve as a comprehensive and a robust MRI-based prognostic marker of disease aggressiveness and survival in solid tumors.

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record.This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article.

Since its renaissance, deep learning has been widely used in various medical imaging tasks and has achieved remarkable success in many medical imaging applications, thereby propelling us into the so-called artificial intelligence (AI) era. It is known that the success of AI is mostly attributed to the availability of big data with annotations for a single task and the advances in high performance computing. However, medical imaging presents unique challenges that confront deep learning approaches. In this survey paper, we first present traits of medical imaging, highlight both clinical needs and technical challenges in medical imaging, and describe how emerging trends in deep learning are addressing these issues. We cover the topics of network architecture, sparse and noisy labels, federating learning, interpretability, uncertainty quantification, etc. Then, we present several case studies that are commonly found in clinical practice, including digital pathology and chest, brain, cardiovascular, and abdominal imaging. Rather than presenting an exhaustive literature survey, we instead describe some prominent research highlights related to these case study applications. We conclude with a discussion and presentation of promising future directions.

Almost 25% of COVID-19 patients end up in ICU needing critical mechanical ventilation support. There is currently no validated objective way to predict which patients will end up needing ventilator support, when the disease is mild and not progressed. N = 869 patients from two sites (D1: N = 822, D2: N = 47) with baseline clinical characteristics and chest CT scans were considered for this study. The entire dataset was randomly divided into 70% training, D1train(N = 606) and 30% test-set (Dtest: D1test(N = 216) + D2(N = 47)). An expert radiologist delineated ground-glass-opacities (GGOs) and consolidation regions on a subset of D1train, (D1train_sub, N = 88). These regions were automatically segmented and used along with their corresponding CT volumes to train an imaging AI predictor (AIP) on D1trainto predict the need of mechanical ventilators for COVID-19 patients. Finally, top five prognostic clinical factors selected using univariate analysis were integrated with AIP to construct an integrated clinical and AI imaging nomogram (ClAIN). Univariate analysis identified lactate dehydrogenase, prothrombin time, aspartate aminotransferase, %lymphocytes, albumin as top five prognostic clinical features. AIP yielded an AUC of 0.81 on Dtestand was independently prognostic irrespective of other clinical parameters on multivariable analysis (ptest. ClAIN outperformed AIP in predicting which COVID-19 patients ended up needing a ventilator. Our results across multiple sites suggest that ClAIN could help identify COVID-19 with severe disease more precisely and likely to end up on a life-saving mechanical ventilation.

Automated nuclear detection is a critical step for a number of computer assisted pathology related image analysis algorithms such as for automated grading of breast cancer tissue specimens. The Nottingham Histologic Score system is highly correlated with the shape and appearance of breast cancer nuclei in histopathological images. However, automated nucleus detection is complicated by 1) the large number of nuclei and the size of high resolution digitized pathology images, and 2) the variability in size, shape, appearance, and texture of the individual nuclei. Recently there has been interest in the application of "Deep Learning" strategies for classification and analysis of big image data. Histopathology, given its size and complexity, represents an excellent use case for application of deep learning strategies. In this paper, a Stacked Sparse Autoencoder (SSAE), an instance of a deep learning strategy, is presented for efficient nuclei detection on high-resolution histopathological images of breast cancer. The SSAE learns high-level features from just pixel intensities alone in order to identify distinguishing features of nuclei. A sliding window operation is applied to each image in order to represent image patches via high-level features obtained via the auto-encoder, which are then subsequently fed to a classifier which categorizes each image patch as nuclear or non-nuclear. Across a cohort of 500 histopathological images (2200 × 2200) and approximately 3500 manually segmented individual nuclei serving as the groundtruth, SSAE was shown to have an improved F-measure 84.49% and an average area under Precision-Recall curve (AveP) 78.83%. The SSAE approach also out-performed nine other state of the art nuclear detection strategies.

Quantitative histomorphometry (QH) refers to the process of computationally modeling disease appearance on digital pathology images by extracting hundreds of image features and using them to predict disease presence or outcome. Since constructing a robust and interpretable classifier is challenging in a high dimensional feature space, dimensionality reduction (DR) is often implemented prior to classifier construction. However, when DR is performed it can be challenging to quantify the contribution of each of the original features to the final classification result. We have previously presented a method for scoring features based on their importance for classification on an embedding derived via principal components analysis (PCA). However, nonlinear DR involves the eigen-decomposition of a kernel matrix rather than the data itself, compounding the issue of classifier interpretability. In this paper we present feature importance in nonlinear embeddings (FINE), an extension of our PCA-based feature scoring method to kernel PCA (KPCA), as well as several NLDR algorithms that can be cast as variants of KPCA. FINE is applied to four digital pathology datasets to identify key QH features for predicting the risk of breast and prostate cancer recurrence. Measures of nuclear and glandular architecture and clusteredness were found to play an important role in predicting the likelihood of recurrence of both breast and prostate cancers. Compared to the t-test, Fisher score, and Gini index, FINE was able to identify a stable set of features that provide good classification accuracy on four publicly available datasets from the NIPS 2003 Feature Selection Challenge.

In this work, we present a new methodology to facilitate prediction of recurrent prostate cancer (CaP) following radical prostatectomy (RP) via the integration of quantitative image features and protein expression in the excised prostate. Creating a fused predictor from high-dimensional data streams is challenging because the classifier must 1) account for the "curse of dimensionality" problem, which hinders classifier performance when the number of features exceeds the number of patient studies and 2) balance potential mismatches in the number of features across different channels to avoid classifier bias towards channels with more features. Our new data integration methodology, supervised Multi-view Canonical Correlation Analysis (sMVCCA), aims to integrate infinite views of highdimensional data to provide more amenable data representations for disease classification. Additionally, we demonstrate sMVCCA using Spearman's rank correlation which, unlike Pearson's correlation, can account for nonlinear correlations and outliers. Forty CaP patients with pathological Gleason scores 6-8 were considered for this study. 21 of these men revealed biochemical recurrence (BCR) following RP, while 19 did not. For each patient, 189 quantitative histomorphometric attributes and 650 protein expression levels were extracted from the primary tumor nodule. The fused histomorphometric/proteomic representation via sMVCCA combined with a random forest classifier predicted BCR with a mean AUC of 0.74 and a maximum AUC of 0.9286. We found sMVCCA to perform statistically significantly (p < 0.05) better than comparative state-of-the-art data fusion strategies for predicting BCR. Furthermore, Kaplan-Meier analysis demonstrated improved BCR-free survival prediction for the sMVCCA-fused classifier as compared to histology or proteomic features alone.

Quantitative measurements of spatial arrangement of nuclei in histopathology images for different cancers has been shown to have prognostic value. Traditionally, graph algorithms (with cell/nuclei as node) have been used to characterize the spatial arrangement of these cells. However, these graphs inherently extract only global features of cell or nuclear architecture and, therefore, important information at the local level may be left unexploited. Additionally, since the graph construction does not draw a distinction between nuclei in the stroma or epithelium, the graph edges often traverse the stromal and epithelial regions. In this paper, we present a new spatially aware cell cluster (SpACC1) graph that can efficiently and accurately model local nuclear interactions, separately within the stromal and epithelial regions alone. SpACC1 is built locally on nodes that are defined on groups/clusters of nuclei rather than individual nuclei. Local nodes are connected with edges which have a certain probability of connectedness. The SpACC1 graph allows for exploration of (a) contribution of nuclear arrangement within the stromal and epithelial regions separately and (b) combined contribution of stromal and epithelial nuclear architecture in predicting disease aggressiveness and patient outcome. In a cohort of 160 p16+ oropharyngeal tumors (141 non-progressors and 19 progressors), a support vector machine (SVM) classifier in conjunction with 7 graph features extracted from the SpACC1 graph yielded a mean accuracy of over 90% with PPV of 89.4% in distinguishing between progressors and non-progressors. Our results suggest that (a) stromal nuclear architecture has a role to play in predicting disease aggressiveness and that (b) combining nuclear architectural contributions from the stromal and epithelial regions yields superior prognostic accuracy compared to individual contributions from stroma and epithelium alone.

BACKGROUND: Automated classification of histopathology involves identification of multiple classes, including benign, cancerous, and confounder categories. The confounder tissue classes can often mimic and share attributes with both the diseased and normal tissue classes, and can be particularly difficult to identify, both manually and by automated classifiers. In the case of prostate cancer, they may be several confounding tissue types present in a biopsy sample, posing as major sources of diagnostic error for pathologists. Two common multi-class approaches are one-shot classification (OSC), where all classes are identified simultaneously, and one-versus-all (OVA), where a "target" class is distinguished from all "non-target" classes. OSC is typically unable to handle discrimination of classes of varying similarity (e.g. with images of prostate atrophy and high grade cancer), while OVA forces several heterogeneous classes into a single "non-target" class. In this work, we present a cascaded (CAS) approach to classifying prostate biopsy tissue samples, where images from different classes are grouped to maximize intra-group homogeneity while maximizing inter-group heterogeneity. RESULTS: We apply the CAS approach to categorize 2000 tissue samples taken from 214 patient studies into seven classes: epithelium, stroma, atrophy, prostatic intraepithelial neoplasia (PIN), and prostate cancer Gleason grades 3, 4, and 5. A series of increasingly granular binary classifiers are used to split the different tissue classes until the images have been categorized into a single unique class. Our automatically-extracted image feature set includes architectural features based on location of the nuclei within the tissue sample as well as texture features extracted on a per-pixel level. The CAS strategy yields a positive predictive value (PPV) of 0.86 in classifying the 2000 tissue images into one of 7 classes, compared with the OVA (0.77 PPV) and OSC approaches (0.76 PPV). CONCLUSIONS: Use of the CAS strategy increases the PPV for a multi-category classification system over two common alternative strategies. In classification problems such as histopathology, where multiple class groups exist with varying degrees of heterogeneity, the CAS system can intelligently assign class labels to objects by performing multiple binary classifications according to domain knowledge.

BACKGROUND: Dimensionality reduction (DR) enables the construction of a lower dimensional space (embedding) from a higher dimensional feature space while preserving object-class discriminability. However several popular DR approaches suffer from sensitivity to choice of parameters and/or presence of noise in the data. In this paper, we present a novel DR technique known as consensus embedding that aims to overcome these problems by generating and combining multiple low-dimensional embeddings, hence exploiting the variance among them in a manner similar to ensemble classifier schemes such as Bagging. We demonstrate theoretical properties of consensus embedding which show that it will result in a single stable embedding solution that preserves information more accurately as compared to any individual embedding (generated via DR schemes such as Principal Component Analysis, Graph Embedding, or Locally Linear Embedding). Intelligent sub-sampling (via mean-shift) and code parallelization are utilized to provide for an efficient implementation of the scheme. RESULTS: Applications of consensus embedding are shown in the context of classification and clustering as applied to: (1) image partitioning of white matter and gray matter on 10 different synthetic brain MRI images corrupted with 18 different combinations of noise and bias field inhomogeneity, (2) classification of 4 high-dimensional gene-expression datasets, (3) cancer detection (at a pixel-level) on 16 image slices obtained from 2 different high-resolution prostate MRI datasets. In over 200 different experiments concerning classification and segmentation of biomedical data, consensus embedding was found to consistently outperform both linear and non-linear DR methods within all applications considered. CONCLUSIONS: We have presented a novel framework termed consensus embedding which leverages ensemble classification theory within dimensionality reduction, allowing for application to a wide range of high-dimensional biomedical data classification and segmentation problems. Our generalizable framework allows for improved representation and classification in the context of both imaging and non-imaging data. The algorithm offers a promising solution to problems that currently plague DR methods, and may allow for extension to other areas of biomedical data analysis.

Active contours and active shape models (ASM) have been widely employed in image segmentation. A major limitation of active contours, however, is in their 1) inability to resolve boundaries of intersecting objects and to 2) handle occlusion. Multiple overlapping objects are typically segmented out as a single object. On the other hand, ASMs are limited by point correspondence issues since object landmarks need to be identified across multiple objects for initial object alignment. ASMs are also are constrained in that they can usually only segment a single object in an image. In this paper, we present a novel synergistic boundary and region-based active contour model that incorporates shape priors in a level set formulation with automated initialization based on watershed. We demonstrate an application of these synergistic active contour models using multiple level sets to segment nuclear and glandular structures on digitized histopathology images of breast and prostate biopsy specimens. Unlike previous related approaches, our model is able to resolve object overlap and separate occluded boundaries of multiple objects simultaneously. The energy functional of the active contour is comprised of three terms. The first term is the prior shape term, modeled on the object of interest, thereby constraining the deformation achievable by the active contour. The second term, a boundary-based term detects object boundaries from image gradients. The third term drives the shape prior and the contour towards the object boundary based on region statistics. The results of qualitative and quantitative evaluation on 100 prostate and 14 breast cancer histology images for the task of detecting and segmenting nuclei and lymphocytes reveals that the model easily outperforms two state of the art segmentation schemes (geodesic active contour and Rousson shape-based model) and on average is able to resolve up to 91% of overlapping/occluded structures in the images.

Active shape models (ASMs) and active appearance models (AAMs) are popular approaches for medical image segmentation that use shape information to drive the segmentation process. Both approaches rely on image derived landmarks (specified either manually or automatically) to define the object's shape, which require accurate triangulation and alignment. An alternative approach to modeling shape is the levelset representation, defined as a set of signed distances to the object's surface. In addition, using multiple image derived attributes (IDAs) such as gradient information has previously shown to offer improved segmentation results when applied to ASMs, yet little work has been done exploring IDAs in the context of AAMs. In this work, we present a novel AAM methodology that utilizes the levelset implementation to overcome the issues relating to specifying landmarks, and locates the object of interest in a new image using a registration based scheme. Additionally, the framework allows for incorporation of multiple IDAs. Our multifeature landmark-free AAM (MFLAAM) utilizes an efficient, intuitive, and accurate algorithm for identifying those IDAs that will offer the most accurate segmentations. In this paper, we evaluate our MFLAAM scheme for the problem of prostate segmentation from T2-w MRI volumes. On a cohort of 108 studies, the levelset MFLAAM yielded a mean Dice accuracy of 88% ± 5%, and a mean surface error of 1.5 mm ±.8 mm with a segmentation time of 150/s per volume. In comparison, a state of the art AAM yielded mean Dice and surface error values of 86% ± 9% and 1.6 mm ± 1.0 mm, respectively. The differences with respect to our levelset-based MFLAAM model are statistically significant . In addition, our results were in most cases superior to several recent state of the art prostate MRI segmentation methods.

BACKGROUND: Supervised classifiers for digital pathology can improve the ability of physicians to detect and diagnose diseases such as cancer. Generating training data for classifiers is problematic, since only domain experts (e.g. pathologists) can correctly label ground truth data. Additionally, digital pathology datasets suffer from the "minority class problem", an issue where the number of exemplars from the non-target class outnumber target class exemplars which can bias the classifier and reduce accuracy. In this paper, we develop a training strategy combining active learning (AL) with class-balancing. AL identifies unlabeled samples that are "informative" (i.e. likely to increase classifier performance) for annotation, avoiding non-informative samples. This yields high accuracy with a smaller training set size compared with random learning (RL). Previous AL methods have not explicitly accounted for the minority class problem in biomedical images. Pre-specifying a target class ratio mitigates the problem of training bias. Finally, we develop a mathematical model to predict the number of annotations (cost) required to achieve balanced training classes. In addition to predicting training cost, the model reveals the theoretical properties of AL in the context of the minority class problem. RESULTS: Using this class-balanced AL training strategy (CBAL), we build a classifier to distinguish cancer from non-cancer regions on digitized prostate histopathology. Our dataset consists of 12,000 image regions sampled from 100 biopsies (58 prostate cancer patients). We compare CBAL against: (1) unbalanced AL (UBAL), which uses AL but ignores class ratio; (2) class-balanced RL (CBRL), which uses RL with a specific class ratio; and (3) unbalanced RL (UBRL). The CBAL-trained classifier yields 2% greater accuracy and 3% higher area under the receiver operating characteristic curve (AUC) than alternatively-trained classifiers. Our cost model accurately predicts the number of annotations necessary to obtain balanced classes. The accuracy of our prediction is verified by empirically-observed costs. Finally, we find that over-sampling the minority class yields a marginal improvement in classifier accuracy but the improved performance comes at the expense of greater annotation cost. CONCLUSIONS: We have combined AL with class balancing to yield a general training strategy applicable to most supervised classification problems where the dataset is expensive to obtain and which suffers from the minority class problem. An intelligent training strategy is a critical component of supervised classification, but the integration of AL and intelligent choice of class ratios, as well as the application of a general cost model, will help researchers to plan the training process more quickly and effectively.

BACKGROUND: Multimodal data, especially imaging and non-imaging data, is being routinely acquired in the context of disease diagnostics; however, computational challenges have limited the ability to quantitatively integrate imaging and non-imaging data channels with different dimensionalities and scales. To the best of our knowledge relatively few attempts have been made to quantitatively fuse such data to construct classifiers and none have attempted to quantitatively combine histology (imaging) and proteomic (non-imaging) measurements for making diagnostic and prognostic predictions. The objective of this work is to create a common subspace to simultaneously accommodate both the imaging and non-imaging data (and hence data corresponding to different scales and dimensionalities), called a metaspace. This metaspace can be used to build a meta-classifier that produces better classification results than a classifier that is based on a single modality alone. Canonical Correlation Analysis (CCA) and Regularized CCA (RCCA) are statistical techniques that extract correlations between two modes of data to construct a homogeneous, uniform representation of heterogeneous data channels. In this paper, we present a novel modification to CCA and RCCA, Supervised Regularized Canonical Correlation Analysis (SRCCA), that (1) enables the quantitative integration of data from multiple modalities using a feature selection scheme, (2) is regularized, and (3) is computationally cheap. We leverage this SRCCA framework towards the fusion of proteomic and histologic image signatures for identifying prostate cancer patients at the risk of 5 year biochemical recurrence following radical prostatectomy. RESULTS: A cohort of 19 grade, stage matched prostate cancer patients, all of whom had radical prostatectomy, including 10 of whom had biochemical recurrence within 5 years of surgery and 9 of whom did not, were considered in this study. The aim was to construct a lower fused dimensional metaspace comprising both the histological and proteomic measurements obtained from the site of the dominant nodule on the surgical specimen. In conjunction with SRCCA, a random forest classifier was able to identify prostate cancer patients, who developed biochemical recurrence within 5 years, with a maximum classification accuracy of 93%. CONCLUSIONS: The classifier performance in the SRCCA space was found to be statistically significantly higher compared to the fused data representations obtained, not only from CCA and RCCA, but also two other statistical techniques called Principal Component Analysis and Partial Least Squares Regression. These results suggest that SRCCA is a computationally efficient and a highly accurate scheme for representing multimodal (histologic and proteomic) data in a metaspace and that it could be used to construct fused biomarkers for predicting disease recurrence and prognosis.

The ability of classification systems to adjust their performance (sensitivity/specificity) is essential for tasks in which certain errors are more significant than others. For example, mislabeling cancerous lesions as benign is typically more detrimental than mislabeling benign lesions as cancerous. Unfortunately, methods for modifying the performance of Markov random field (MRF) based classifiers are noticeably absent from the literature, and thus most such systems restrict their performance to a single, static operating point (a paired sensitivity/specificity). To address this deficiency we present weighted maximum posterior marginals (WMPM) estimation, an extension of maximum posterior marginals (MPM) estimation. Whereas the MPM cost function penalizes each error equally, the WMPM cost function allows misclassifications associated with certain classes to be weighted more heavily than others. This creates a preference for specific classes, and consequently a means for adjusting classifier performance. Realizing WMPM estimation (like MPM estimation) requires estimates of the posterior marginal distributions. The most prevalent means for estimating these--proposed by Marroquin--utilizes a Markov chain Monte Carlo (MCMC) method. Though Marroquin's method (M-MCMC) yields estimates that are sufficiently accurate for MPM estimation, they are inadequate for WMPM. To more accurately estimate the posterior marginals we present an equally simple, but more effective extension of the MCMC method (E-MCMC). Assuming an identical number of iterations, E-MCMC as compared to M-MCMC yields estimates with higher fidelity, thereby 1) allowing a far greater number and diversity of operating points and 2) improving overall classifier performance. To illustrate the utility of WMPM and compare the efficacies of M-MCMC and E-MCMC, we integrate them into our MRF-based classification system for detecting cancerous glands in (whole-mount or quarter) histological sections of the prostate.

The presence of lymphocytic infiltration (LI) has been correlated with nodal metastasis and tumor recurrence in HER2+breast cancer (BC), making it important to study LI. The ability to detect and quantify extent of LI could serve as an image based prognostic tool for HER2+ BC patients. Lymphocyte segmentation in H & E-stained BC histopathology images is, however, complicated due to the similarity in appearance between lymphocyte nuclei and cancer nuclei. Additional challenges include biological variability, histological artifacts, and high prevalence of overlapping objects. Although active contours are widely employed in segmentation, they are limited in their ability to segment overlapping objects. In this paper, we propose a segmentation scheme (EMaGACOR) that integrates Expectation Maximization (EM) based segmentation with a geodesic active contour (GAC). Additionally, a novel heuristic edge-path algorithm exploits the size of lymphocytes to split contours that enclose overlapping objects. For a total of 62 HER2+ breast biopsy images, EMaGACOR was found to have a detection sensitivity of over 90% and a positive predictive value of over 78%. By comparison, EMaGAC (model without overlap resolution) and GAC (Randomly initialized geodesic active contour) model yielded corresponding sensitivities of 57.4% and 26.7%, respectively. Furthermore, EMaGACOR was able to resolve over 92% of overlaps. Our scheme was found to be robust, reproducible, accurate, and could potentially be applied to other biomedical image segmentation applications.

The recent explosion in procurement and availability of high-dimensional gene and protein expression profile data sets for cancer diagnostics has necessitated the development of sophisticated machine learning tools with which to analyze them. While some investigators are focused on identifying informative genes and proteins that play a role in specific diseases, other researchers have attempted instead to use patients based on their expression profiles to prognosticate disease status. A major limitation in the ability to accurately classify these high-dimensional data sets stems from the "curse of dimensionality," occurring in situations where the number of genes or peptides significantly exceeds the total number of patient samples. Previous attempts at dealing with this issue have mostly centered on the use of a dimensionality reduction (DR) scheme, principal component analysis (PCA), to obtain a low-dimensional projection of the high-dimensional data. However, linear PCA and other linear DR methods, which rely on euclidean distances to estimate object similarity, do not account for the inherent underlying nonlinear structure associated with most biomedical data. While some researchers have begun to explore nonlinear DR methods for computer vision problems such as face detection and recognition, to the best of our knowledge, few such attempts have been made for classification and visualization of high-dimensional biomedical data. The motivation behind this work is to identify the appropriate DR methods for analysis of high-dimensional gene and protein expression studies. Toward this end, we empirically and rigorously compare three nonlinear (Isomap, Locally Linear Embedding, and Laplacian Eigenmaps) and three linear DR schemes (PCA, Linear Discriminant Analysis, and Multidimensional Scaling) with the intent of determining a reduced subspace representation in which the individual object classes are more easily discriminable. Owing to the inherent nonlinear structure of gene and protein expression studies, our claim is that the nonlinear DR methods provide a more truthful low-dimensional representation of the data compared to the linear DR schemes. Evaluation of the DR schemes was done by 1) assessing the discriminability of two supervised classifiers (Support Vector Machine and C4.5 Decision Trees) in the different low- dimensional data embeddings and 2) five cluster validity measures to evaluate the size, distance, and tightness of object aggregates in the low-dimensional space. For each of the seven evaluation measures considered, statistically significant improvement in the quality of the embeddings across 10 cancer data sets via the use of three nonlinear DR schemes over three linear DR techniques was observed. Similar trends were observed when linear and nonlinear DR was applied to the high-dimensional data following feature pruning to isolate the most informative features. Qualitative evaluation of the low-dimensional data embedding obtained via the six DR methods further suggests that the nonlinear schemes are better able to identify potential novel classes (e.g., cancer subtypes) within the data.

Prostatic adenocarcinoma is the most commonly occurring cancer among men in the United States, second only to skin cancer. Currently, the only definitive method to ascertain the presence of prostatic cancer is by trans-rectal ultrasound (TRUS) directed biopsy. Owing to the poor image quality of ultrasound, the accuracy of TRUS is only 20%-25%. High-resolution magnetic resonance imaging (MRI) has been shown to have a higher accuracy of prostate cancer detection compared to ultrasound. Consequently, several researchers have been exploring the use of high resolution MRI in performing prostate biopsies. Visual detection of prostate cancer, however, continues to be difficult owing to its apparent lack of shape, and the fact that several malignant and benign structures have overlapping intensity and texture characteristics. In this paper, we present a fully automated computer-aided detection (CAD) system for detecting prostatic adenocarcinoma from 4 Tesla ex vivo magnetic resonance (MR) imagery of the prostate. After the acquired MR images have been corrected for background inhomogeneity and nonstandardness, novel three-dimensional (3-D) texture features are extracted from the 3-D MRI scene. A Bayesian classifier then assigns each image voxel a "likelihood" of malignancy for each feature independently. The "likelihood" images generated in this fashion are then combined using an optimally weighted feature combination scheme. Quantitative evaluation was performed by comparing the CAD results with the manually ascertained ground truth for the tumor on the MRI. The tumor labels on the MR slices were determined manually by an expert by visually registering the MR slices with the corresponding regions on the histology slices. We evaluated our CAD system on a total of 33 two-dimensional (2-D) MR slices from five different 3-D MR prostate studies. Five slices from two different glands were used for training. Our feature combination scheme was found to outperform the individual texture features, and also other popularly used feature combination methods, including AdaBoost, ensemble averaging, and majority voting. Further, in several instances our CAD system performed better than the experts in terms of accuracy, the expert segmentations being determined solely from visual inspection of the MRI data. In addition, the intrasystem variability (changes in CAD accuracy with changes in values of system parameters) was significantly lower than the corresponding intraobserver and interobserver variability. CAD performance was found to be very similar for different training sets. Future work will focus on extending the methodology to guide high-resolution MRI-assisted in vivo prostate biopsies.

Image intensity standardization is a postprocessing method designed for correcting acquisition-to-acquisition signal intensity variations (nonstandardness) inherent in magnetic resonance (MR) images. Inhomogeneity correction is a process used to suppress the low frequency background nonuniformities (inhomogeneities) of the image domain that exist in MR images. Both these procedures have important implications in MR image analysis. The effects of these postprocessing operations on improvement of image quality in isolation has been well documented. However, the combined effects of these two processes on MR images and how the processes influence each other have not been studied thus far. In this paper, we evaluate the effect of inhomogeneity correction followed by standardization and vice-versa on MR images in order to determine the best sequence to follow for enhancing image quality. We conducted experiments on several clinical and phantom data sets (nearly 4000 three-dimensional MR images were analyzed) corresponding to four different MRI protocols. Different levels of artificial nonstandardness, and different models and levels of artificial background inhomogeneity were used in these experiments. Our results indicate that improved standardization can be achieved by preceding it with inhomogeneity correction. There is no statistically significant difference in image quality obtained between the results of standardization followed by correction and that of correction followed by standardization from the perspective of inhomogeneity correction. The correction operation is found to bias the effect of standardization. We demonstrate this bias both qualitatively and quantitatively by using two different methods of inhomogeneity correction. We also show that this bias in standardization is independent of the specific inhomogeneity correction method used. The effect of this bias due to correction was also seen in magnetization transfer ratio (MTR) images, which are naturally endowed with the standardness property. Standardization, on the other hand, does not seem to influence the correction operation. It is also found that longer sequences of repeated correction and standardization operations do not considerably improve image quality. These results were found to hold for the clinical and the phantom data sets, for different MRI protocols, for different levels of artificial nonstandardness, for different models and levels of artificial inhomogeneity, for different correction methods, and for images that were endowed with inherent standardness as well as for those that were standardized by using the intensity standardization method. Overall, we conclude that inhomogeneity correction followed by intensity standardization is the best sequence to follow from the perspective of both image quality and computational efficiency.

Breast cancer is the most frequently diagnosed malignancy and the second leading cause of mortality in women. In the last decade, ultrasound along with digital mammography has come to be regarded as the gold standard for breast cancer diagnosis. Automatically detecting tumors and extracting lesion boundaries in ultrasound images is difficult due to their specular nature and the variance in shape and appearance of sonographic lesions. Past work on automated ultrasonic breast lesion segmentation has not addressed important issues such as shadowing artifacts or dealing with similar tumor like structures in the sonogram. Algorithms that claim to automatically classify ultrasonic breast lesions, rely on manual delineation of the tumor boundaries. In this paper, we present a novel technique to automatically find lesion margins in ultrasound images, by combining intensity and texture with empirical domain specific knowledge along with directional gradient and a deformable shape-based model. The images are first filtered to remove speckle noise and then contrast enhanced to emphasize the tumor regions. For the first time, a mathematical formulation of the empirical rules used by radiologists in detecting ultrasonic breast lesions, popularly known as the "Stavros Criteria" is presented in this paper. We have applied this formulation to automatically determine a seed point within the image. Probabilistic classification of image pixels based on intensity and texture is followed by region growing using the automatically determined seed point to obtain an initial segmentation of the lesion. Boundary points are found on the directional gradient of the image. Outliers are removed by a process of recursive refinement. These boundary points are then supplied as an initial estimate to a deformable model. Incorporating empirical domain specific knowledge along with low and high-level knowledge makes it possible to avoid shadowing artifacts and lowers the chance of confusing similar tumor like structures for the lesion. The system was validated on a database of breast sonograms for 42 patients. The average mean boundary error between manual and automated segmentation was 6.6 pixels and the normalized true positive area overlap was 75.1%. The algorithm was found to be robust to 1) variations in system parameters, 2) number of training samples used, and 3) the position of the seed point within the tumor. Running time for segmenting a single sonogram was 18 s on a 1.8-GHz Pentium machine.

This paper presents a methodology for automatically identifying human actions in either the frontal or the lateral view. By tracking the movement of the head of the subject over successive frames of a monocular grayscale image sequence, we recognize 12 different actions. The head is segmented automatically in each frame, and the feature vectors extracted. Input sequences captured from a fixed CCD camera are matched against stored models of actions. The system uses the nearest neighbor classifier to identify the test action.