A Machine Learning Approach Enables Quantitative Measurement of Liver Histology and Disease Monitoring in NASH.

BACKGROUND AND AIMS: Manual histological assessment is currently the accepted standard for diagnosing and monitoring disease progression in NASH, but is limited by variability in interpretation and insensitivity to change. Thus, there is a critical need for improved tools to assess liver pathology in order to risk stratify NASH patients and monitor treatment response. APPROACH AND RESULTS: Here, we describe a machine learning (ML)-based approach to liver histology assessment, which accurately characterizes disease severity and heterogeneity, and sensitively quantifies treatment response in NASH. We use samples from three randomized controlled trials to build and then validate deep convolutional neural networks to measure key histological features in NASH, including steatosis, inflammation, hepatocellular ballooning, and fibrosis. The ML-based predictions showed strong correlations with expert pathologists and were prognostic of progression to cirrhosis and liver-related clinical events. We developed a heterogeneity-sensitive metric of fibrosis response, the Deep Learning Treatment Assessment Liver Fibrosis score, which measured antifibrotic treatment effects that went undetected by manual pathological staging and was concordant with histological disease progression. CONCLUSIONS: Our ML method has shown reproducibility and sensitivity and was prognostic for disease progression, demonstrating the power of ML to advance our understanding of disease heterogeneity in NASH, risk stratify affected patients, and facilitate the development of therapies.

Highly multiplexed single-cell analysis of formalin-fixed, paraffin-embedded cancer tissue.

Limitations on the number of unique protein and DNA molecules that can be characterized microscopically in a single tissue specimen impede advances in understanding the biological basis of health and disease. Here we present a multiplexed fluorescence microscopy method (MxIF) for quantitative, single-cell, and subcellular characterization of multiple analytes in formalin-fixed paraffin-embedded tissue. Chemical inactivation of fluorescent dyes after each image acquisition round allows reuse of common dyes in iterative staining and imaging cycles. The mild inactivation chemistry is compatible with total and phosphoprotein detection, as well as DNA FISH. Accurate computational registration of sequential images is achieved by aligning nuclear counterstain-derived fiducial points. Individual cells, plasma membrane, cytoplasm, nucleus, tumor, and stromal regions are segmented to achieve cellular and subcellular quantification of multiplexed targets. In a comparison of pathologist scoring of diaminobenzidine staining of serial sections and automated MxIF scoring of a single section, human epidermal growth factor receptor 2, estrogen receptor, p53, and androgen receptor staining by diaminobenzidine and MxIF methods yielded similar results. Single-cell staining patterns of 61 protein antigens by MxIF in 747 colorectal cancer subjects reveals extensive tumor heterogeneity, and cluster analysis of divergent signaling through ERK1/2, S6 kinase 1, and 4E binding protein 1 provides insights into the spatial organization of mechanistic target of rapamycin and MAPK signal transduction. Our results suggest MxIF should be broadly applicable to problems in the fields of basic biological research, drug discovery and development, and clinical diagnostics.

AI powered quantification of nuclear morphology in cancers enables prediction of genome instability and prognosis.

While alterations in nucleus size, shape, and color are ubiquitous in cancer, comprehensive quantification of nuclear morphology across a whole-slide histologic image remains a challenge. Here, we describe the development of a pan-tissue, deep learning-based digital pathology pipeline for exhaustive nucleus detection, segmentation, and classification and the utility of this pipeline for nuclear morphologic biomarker discovery. Manually-collected nucleus annotations were used to train an object detection and segmentation model for identifying nuclei, which was deployed to segment nuclei in H&E-stained slides from the BRCA, LUAD, and PRAD TCGA cohorts. Interpretable features describing the shape, size, color, and texture of each nucleus were extracted from segmented nuclei and compared to measurements of genomic instability, gene expression, and prognosis. The nuclear segmentation and classification model trained herein performed comparably to previously reported models. Features extracted from the model revealed differences sufficient to distinguish between BRCA, LUAD, and PRAD. Furthermore, cancer cell nuclear area was associated with increased aneuploidy score and homologous recombination deficiency. In BRCA, increased fibroblast nuclear area was indicative of poor progression-free and overall survival and was associated with gene expression signatures related to extracellular matrix remodeling and anti-tumor immunity. Thus, we developed a powerful pan-tissue approach for nucleus segmentation and featurization, enabling the construction of predictive models and the identification of features linking nuclear morphology with clinically-relevant prognostic biomarkers across multiple cancer types.

AI-based automation of enrollment criteria and endpoint assessment in clinical trials in liver diseases.

Clinical trials in metabolic dysfunction-associated steatohepatitis (MASH, formerly known as nonalcoholic steatohepatitis) require histologic scoring for assessment of inclusion criteria and endpoints. However, variability in interpretation has impacted clinical trial outcomes. We developed an artificial intelligence-based measurement (AIM) tool for scoring MASH histology (AIM-MASH). AIM-MASH predictions for MASH Clinical Research Network necroinflammation grades and fibrosis stages were reproducible (κ = 1) and aligned with expert pathologist consensus scores (κ = 0.62-0.74). The AIM-MASH versus consensus agreements were comparable to average pathologists for MASH Clinical Research Network scores (82% versus 81%) and fibrosis (97% versus 96%). Continuous scores produced by AIM-MASH for key histological features of MASH correlated with mean pathologist scores and noninvasive biomarkers and strongly predicted progression-free survival in patients with stage 3 (P < 0.0001) and stage 4 (P = 0.03) fibrosis. In a retrospective analysis of the ATLAS trial (NCT03449446), responders receiving study treatment showed a greater continuous change in fibrosis compared with placebo (P = 0.02). Overall, these results suggest that AIM-MASH may assist pathologists in histologic review of MASH clinical trials, reducing inter-rater variability on trial outcomes and offering a more sensitive and reproducible measure of patient responses.

Integration of deep learning-based histopathology and transcriptomics reveals key genes associated with fibrogenesis in patients with advanced NASH.

Nonalcoholic steatohepatitis (NASH) is the most common chronic liver disease globally and a leading cause for liver transplantation in the US. Its pathogenesis remains imprecisely defined. We combined two high-resolution modalities to tissue samples from NASH clinical trials, machine learning (ML)-based quantification of histological features and transcriptomics, to identify genes that are associated with disease progression and clinical events. A histopathology-driven 5-gene expression signature predicted disease progression and clinical events in patients with NASH with F3 (pre-cirrhotic) and F4 (cirrhotic) fibrosis. Notably, the Notch signaling pathway and genes implicated in liver-related diseases were enriched in this expression signature. In a validation cohort where pharmacologic intervention improved disease histology, multiple Notch signaling components were suppressed.

AI-based histologic scoring enables automated and reproducible assessment of enrollment criteria and endpoints in NASH clinical trials.

Clinical trials in nonalcoholic steatohepatitis (NASH) require histologic scoring for assessment of inclusion criteria and endpoints. However, guidelines for scoring key features have led to variability in interpretation, impacting clinical trial outcomes. We developed an artificial intelligence (AI)-based measurement (AIM) tool for scoring NASH histology (AIM-NASH). AIM-NASH predictions for NASH Clinical Research Network (CRN) grades of necroinflammation and stages of fibrosis aligned with expert consensus scores and were reproducible. Continuous scores produced by AIM-NASH for key histological features of NASH correlated with mean pathologist scores and with noninvasive biomarkers and strongly predicted patient outcomes. In a retrospective analysis of the ATLAS trial, previously unmet pathological endpoints were met when scored by the AIM-NASH algorithm alone. Overall, these results suggest that AIM-NASH may assist pathologists in histologic review of NASH clinical trials, reducing inter-rater variability on trial outcomes and offering a more sensitive and reproducible measure of patient therapeutic response.

Quantitative Image Analysis for Tissue Biomarker Use: A White Paper From the Digital Pathology Association.

Tissue biomarkers have been of increasing utility for scientific research, diagnosing disease, and treatment response prediction. There has been a steady shift away from qualitative assessment toward providing more quantitative scores for these biomarkers. The application of quantitative image analysis has thus become an indispensable tool for in-depth tissue biomarker interrogation in these contexts. This white paper reviews current technologies being employed for quantitative image analysis, their application and pitfalls, regulatory framework demands, and guidelines established for promoting their safe adoption in clinical practice.

Human-interpretable image features derived from densely mapped cancer pathology slides predict diverse molecular phenotypes.

Computational methods have made substantial progress in improving the accuracy and throughput of pathology workflows for diagnostic, prognostic, and genomic prediction. Still, lack of interpretability remains a significant barrier to clinical integration. We present an approach for predicting clinically-relevant molecular phenotypes from whole-slide histopathology images using human-interpretable image features (HIFs). Our method leverages >1.6 million annotations from board-certified pathologists across >5700 samples to train deep learning models for cell and tissue classification that can exhaustively map whole-slide images at two and four micron-resolution. Cell- and tissue-type model outputs are combined into 607 HIFs that quantify specific and biologically-relevant characteristics across five cancer types. We demonstrate that these HIFs correlate with well-known markers of the tumor microenvironment and can predict diverse molecular signatures (AUROC 0.601-0.864), including expression of four immune checkpoint proteins and homologous recombination deficiency, with performance comparable to 'black-box' methods. Our HIF-based approach provides a comprehensive, quantitative, and interpretable window into the composition and spatial architecture of the tumor microenvironment.

The relative distribution of membranous and cytoplasmic met is a prognostic indicator in stage I and II colon cancer.

PURPOSE: The association hepatocyte growth factor receptor (Met) tyrosine kinase with prognosis and survival in colon cancer is unclear, due in part to the limitation of detection methods used. In particular, conventional chromagenic immunohistochemistry (IHC) has several limitations including the inability to separate compartmental measurements. Measurement of membrane, cytoplasm, and nuclear levels of Met could offer a superior approach to traditional IHC. EXPERIMENTAL DESIGN: Fluorescent-based IHC for Met was done in 583 colon cancer patients in a tissue microarray format. Using curvature and intensity-based image analysis, the membrane, nuclear, and cytoplasm were segmented. Probability distributions of Met within each compartment were determined, and an automated scoring algorithm was generated. An optimal score cutpoint was calculated using 500-fold crossvalidation of a training and test data set. For comparison with conventional IHC, a second array from the same tissue microarray block was 3,3'-diaminobenzidine immunostained for Met. RESULTS: In crossvalidated and univariate Cox analysis, the membrane relative to cytoplasm Met score was a significant predictor of survival in stage I (hazard ratio, 0.16; P = 0.006) and in stage II patients (hazard ratio, 0.34; P < or = 0.0005). Similar results were found with multivariate analysis. Met in the membrane alone was not a significant predictor of outcome in all patients or within stage. In the 3,3'-diaminobenzidine-stained array, no associations were found with Met expression and survival. CONCLUSIONS: These data indicate that the relative subcellular distribution of Met, as measured by novel automated image analysis, may be a valuable biomarker for estimating colon cancer prognosis.

Fibrillar and oligomeric beta-amyloid as distinct local biomarkers for Alzheimer's disease.

Beta-amyloid is a key component of Alzheimer's disease (AD) pathology. Researchers in both academic and industry are actively pursuing the development of imaging tracers and techniques to noninvasively measure local levels of beta-amyloid in the Alzheimer's brain. This presentation summarizes recent data and discusses the opportunities and challenges of imaging plaques containing fibrillar beta-amyloid for the early diagnosis and therapeutic monitoring of amyloid targeted therapies. Further, the value and feasibility of measuring the recently described soluble oligomeric form of beta-amyloid as an alternative noninvasive biomarker is also discussed.

Hypoxia-inducible factor-1-dependent regulation of the multidrug resistance (MDR1) gene.

The microenvironment of rapidly growing tumors is associated with increased energy demand and diminished vascular supply, resulting in focal areas of prominent hypoxia. A number of hypoxia-responsive genes have been associated with growing tumors, and here we demonstrate that the multidrug resistance (MDR1) gene product P-glycoprotein, a Mr approximately 170,000 transmembrane protein associated with tumor resistance to chemotherapeutics, is induced by ambient hypoxia. Initial studies using quantitative microarray analysis of RNA revealed an approximately 7-fold increase in MDR in epithelial cells exposed to hypoxia (pO(2) 20 torr, 18 h). These findings were further confirmed at the mRNA and protein level. P-Glycoprotein function was studied by analysis of verapamil-inhibitable efflux of digoxin and rhodamine 123 in intact T84 cells and revealed that hypoxia enhances P-glycoprotein function by as much as 7 +/- 0.4-fold over normoxia. Subsequent studies confirmed hypoxia-elicited MDR1 gene induction and increased P-glycoprotein expression in nontransformed, primary cultures of human microvascular endothelial cells, and analysis of multicellular spheroids subjected to hypoxia revealed increased resistance to doxorubicin. Examination of the MDR1 gene identified a binding site for hypoxia inducible factor-1 (HIF-1), and inhibition of HIF-1 expression by antisense oligonucleotides resulted in significant inhibition of hypoxia-inducible MDR1 expression and a nearly complete loss of basal MDR1 expression. Studies using luciferase promoter constructs revealed a significant increase in activity in cells subjected to hypoxia, and such hypoxia inducibility was lost in truncated constructs lacking the HIF-1 site and in HIF-1 binding site mutants. Extensions of these studies also identified a role for Sp1 in this hypoxia response. Taken together, these data indicate that the MDR1 gene is hypoxia responsive, and such results may identify hypoxia-elicited P-glycoprotein expression as a pathway for resistance of some tumors to chemotherapeutics.

An industry perspective: An update on the adoption of whole slide imaging.

This manuscript is an adaptation of the closing keynote presentation of the Digital Pathology Association Pathology Visions Conference 2015 in Boston, MA, USA. In this presentation, analogies are drawn between the adoption of whole slide imaging (WSI) and other mainstream digital technologies, including digital music and books. In doing so, it is revealed that the adoption of seemingly similar digital technologies does not follow the same adoption profiles and that understanding the unique aspects of value for each customer segment is critical. Finally, a call to action is given to academia and industry to study the value that WSI brings to the global healthcare community.

A multisite validation of whole slide imaging for primary diagnosis using standardized data collection and analysis.

CONTEXT: Text-based reporting and manual arbitration for whole slide imaging (WSI) validation studies are labor intensive and do not allow for consistent, scalable, and repeatable data collection or analysis. OBJECTIVE: The objective of this study was to establish a method of data capture and analysis using standardized codified checklists and predetermined synoptic discordance tables and to use these methods in a pilot multisite validation study. METHODS AND STUDY DESIGN: Fifteen case report form checklists were generated from the College of American Pathology cancer protocols. Prior to data collection, all hypothetical pairwise comparisons were generated, and a level of harm was determined for each possible discordance. Four sites with four pathologists each generated 264 independent reads of 33 cases. Preestablished discordance tables were applied to determine site by site and pooled accuracy, intrareader/intramodality, and interreader intramodality error rates. RESULTS: Over 10,000 hypothetical pairwise comparisons were evaluated and assigned harm in discordance tables. The average difference in error rates between WSI and glass, as compared to ground truth, was 0.75% with a lower bound of 3.23% (95% confidence interval). Major discordances occurred on challenging cases, regardless of modality. The average inter-reader agreement across sites for glass was 76.5% (weighted kappa of 0.68) and for digital it was 79.1% (weighted kappa of 0.72). CONCLUSION: These results demonstrate the feasibility and utility of employing standardized synoptic checklists and predetermined discordance tables to gather consistent, comprehensive diagnostic data for WSI validation studies. This method of data capture and analysis can be applied in large-scale multisite WSI validations.

A computational positron emission tomography simulation model for imaging beta-amyloid in mice.

PURPOSE: We aimed to develop a computational simulation model for beta-amyloid (Abeta) positron emission tomography (PET) imaging. PROCEDURES: Model parameters were set to reproduce levels of Abeta within the PDAPP mouse. Pharmacokinetic curves of virtual tracers were computed and a PET detector simulator was configured for a commercially available preclinical PET-imaging system. RESULTS: We modeled the effects of Abeta therapy and tracer affinity on the ability to differentiate Abeta levels by PET. Varying affinity had a significant effect on the ability to quantitate Abeta. Further, PET tracers for Abeta monomers were more sensitive to the therapeutic reduction in Abeta levels than total brain amyloid. Following therapy, the decrease in total brain Abeta corresponded to the slow rate of change in total amyloid load as expected. CONCLUSIONS: We have developed a first proof-of-concept Abeta-PET simulation model that will be a useful tool in the interpretation of preclinical Abeta imaging data and tracer development.

Murine model of gastrointestinal ischemia associated with complement-dependent injury.

Gastrointestinal ischemia-reperfusion (I/R) injury is often associated with remote tissue injury. Complement activation plays an important role in local and remote tissue injury associated with gastrointestinal I/R. We developed a new murine model of gastrointestinal I/R that has complement-dependent local and remote tissue injury. Twenty, but not thirty, minutes of gastrointestinal ischemia followed by 3 h of reperfusion induced a significant loss of intestinal lactate dehydrogenase that was significantly prevented by a murine anti-murine C5 monoclonal antibody. Anti-C5 also significantly decreased neutrophil infiltration into the gut and lung. Gastrointestinal I/R significantly increased pulmonary intercellular adhesion molecule-1 mRNA and protein expression that was significantly inhibited by anti-C5. Pulmonary macrophage inflammatory protein-2 mRNA was significantly induced by gastrointestinal I/R and inhibited by anti-C5 treatment. These data demonstrate that brief periods of murine gastrointestinal I/R activate complement, leading to tissue injury and neutrophil accumulation. Anti-C5 treatment attenuates tissue injury, neutrophil recruitment, and leukocyte adherence molecule and chemokine expression in the mouse. This model will be well suited to investigate the role of complement-mediated tissue injury and gene expression after gastrointestinal I/R.

Can digital pathology result in cost savings? A financial projection for digital pathology implementation at a large integrated health care organization.

BACKGROUND: Digital pathology offers potential improvements in workflow and interpretive accuracy. Although currently digital pathology is commonly used for research and education, its clinical use has been limited to niche applications such as frozen sections and remote second opinion consultations. This is mainly due to regulatory hurdles, but also to a dearth of data supporting a positive economic cost-benefit. Large scale adoption of digital pathology and the integration of digital slides into the routine anatomic/surgical pathology "slide less" clinical workflow will occur only if digital pathology will offer a quantifiable benefit, which could come in the form of more efficient and/or higher quality care. AIM: As a large academic-based health care organization expecting to adopt digital pathology for primary diagnosis upon its regulatory approval, our institution estimated potential operational cost savings offered by the implementation of an enterprise-wide digital pathology system (DPS). METHODS: Projected cost savings were calculated for the first 5 years following implementation of a DPS based on operational data collected from the pathology department. Projected savings were based on two factors: (1) Productivity and lab consolidation savings; and (2) avoided treatment costs due to improvements in the accuracy of cancer diagnoses among nonsubspecialty pathologists. Detailed analyses of incremental treatment costs due to interpretive errors, resulting in either a false positive or false negative diagnosis, was performed for melanoma and breast cancer and extrapolated to 10 other common cancers. RESULTS: When phased in over 5-years, total cost savings based on anticipated improvements in pathology productivity and histology lab consolidation were estimated at $12.4 million for an institution with 219,000 annual accessions. The main contributing factors to these savings were gains in pathologist clinical full-time equivalent capacity impacted by improved pathologist productivity and workload distribution. Expanding the current localized specialty sign-out model to an enterprise-wide shared general/subspecialist sign-out model could potentially reduce costs of incorrect treatment by $5.4 million. These calculations were based on annual over and under treatment costs for breast cancer and melanoma estimated to be approximately $26,000 and $11,000/case, respectively, and extrapolated to $21,500/case for other cancer types. CONCLUSIONS: The projected 5-year total cost savings for our large academic-based health care organization upon fully implementing a DPS was approximately $18 million. If the costs of digital pathology acquisition and implementation do not exceed this value, the return on investment becomes attractive to hospital administrators. Furthermore, improved patient outcome enabled by this technology strengthens the argument supporting adoption of an enterprise-wide DPS.

The accuracy of dynamic predictive autofocusing for whole slide imaging.

CONTEXT: Whole slide imaging (WSI) for digital pathology involves the rapid automated acquisition of multiple high-power fields from a microscope slide containing a tissue specimen. Capturing each field in the correct focal plane is essential to create high-quality digital images. Others have described a novel focusing method which reduces the number of focal planes required to generate accurate focus. However, this method was not applied dynamically in an automated WSI system under continuous motion. AIMS: This report measures the accuracy of this method when applied in a rapid continuous scan mode using a dual sensor WSI system with interleaved acquisition of images. METHODS: We acquired over 400 tiles in a "stop and go" scan mode, surveying the entire z depth in each tile and used this as ground truth. We compared this ground truth focal height to the focal height determined using a rapid 3-point focus algorithm applied dynamically in a continuous scanning mode. RESULTS: Our data showed the average focal height error of 0.30 (±0.27) µm compared to ground truth, which is well within the system's depth of field. On a tile by tile assessment, approximately 95% of the tiles were within the system's depth of field. Further, this method was six times faster than acquiring tiles compared to the same method in a non-continuous scan mode. CONCLUSIONS: The data indicates that the method employed can yield a significant improvement in scan speed while maintaining highly accurate autofocusing.

Autofocus methods of whole slide imaging systems and the introduction of a second-generation independent dual sensor scanning method.

Accurate focusing is a critical challenge of whole slide imaging, primarily due to inherent tissue topography variability. Traditional line scanning and tile-based scanning systems are limited in their ability to acquire a high degree of focus points while still maintaining high throughput. This review examines limitations with first-generation whole slide scanning systems and explores a novel approach that employs continuous autofocus, referred to as independent dual sensor scanning. This "second-generation" concept decouples image acquisition from focusing, allowing for rapid scanning while maintaining continuous accurate focus. The technical concepts, merits, and limitations of this method are explained and compared to that of a traditional whole slide scanning system.