3D genomic mapping reveals multifocality of human pancreatic precancers.

Pancreatic intraepithelial neoplasias (PanINs) are the most common precursors of pancreatic cancer, but their small size and inaccessibility in humans make them challenging to study1. Critically, the number, dimensions and connectivity of human PanINs remain largely unknown, precluding important insights into early cancer development. Here, we provide a microanatomical survey of human PanINs by analysing 46 large samples of grossly normal human pancreas with a machine-learning pipeline for quantitative 3D histological reconstruction at single-cell resolution. To elucidate genetic relationships between and within PanINs, we developed a workflow in which 3D modelling guides multi-region microdissection and targeted and whole-exome sequencing. From these samples, we calculated a mean burden of 13 PanINs per cm3 and extrapolated that the normal intact adult pancreas harbours hundreds of PanINs, almost all with oncogenic KRAS hotspot mutations. We found that most PanINs originate as independent clones with distinct somatic mutation profiles. Some spatially continuous PanINs were found to contain multiple KRAS mutations; computational and in situ analyses demonstrated that different KRAS mutations localize to distinct cell subpopulations within these neoplasms, indicating their polyclonal origins. The extensive multifocality and genetic heterogeneity of PanINs raises important questions about mechanisms that drive precancer initiation and confer differential progression risk in the human pancreas. This detailed 3D genomic mapping of molecular alterations in human PanINs provides an empirical foundation for early detection and rational interception of pancreatic cancer.

CODA: quantitative 3D reconstruction of large tissues at cellular resolution.

A central challenge in biology is obtaining high-content, high-resolution information while analyzing tissue samples at volumes relevant to disease progression. We address this here with CODA, a method to reconstruct exceptionally large (up to multicentimeter cubed) tissues at subcellular resolution using serially sectioned hematoxylin and eosin-stained tissue sections. Here we demonstrate CODA's ability to reconstruct three-dimensional (3D) distinct microanatomical structures in pancreas, skin, lung and liver tissues. CODA allows creation of readily quantifiable tissue volumes amenable to biological research. As a testbed, we assess the microanatomy of the human pancreas during tumorigenesis within the branching pancreatic ductal system, labeling ten distinct structures to examine heterogeneity and structural transformation during neoplastic progression. We show that pancreatic precancerous lesions develop into distinct 3D morphological phenotypes and that pancreatic cancer tends to spread far from the bulk tumor along collagen fibers that are highly aligned to the 3D curves of ductal, lobular, vascular and neural structures. Thus, CODA establishes a means to transform broadly the structural study of human diseases through exploration of exhaustively labeled 3D microarchitecture.

Interactive Multimedia Reporting Technical Considerations: HIMSS-SIIM Collaborative White Paper.

Despite technological advances in the analysis of digital images for medical consultations, many health information systems lack the ability to correlate textual descriptions of image findings linked to the actual images. Images and reports often reside in separate silos in the medical record throughout the process of image viewing, report authoring, and report consumption. Forward-thinking centers and early adopters have created interactive reports with multimedia elements and embedded hyperlinks in reports that connect the narrative text with the related source images and measurements. Most of these solutions rely on proprietary single-vendor systems for viewing and reporting in the absence of any encompassing industry standards to facilitate interoperability with the electronic health record (EHR) and other systems. International standards have enabled the digitization of image acquisition, storage, viewing, and structured reporting. These provide the foundation to discuss enhanced reporting. Lessons learned in the digital transformation of radiology and pathology can serve as a basis for interactive multimedia reporting (IMR) across image-centric medical specialties. This paper describes the standard-based infrastructure and communications to fulfill recently defined clinical requirements through a consensus from an international workgroup of multidisciplinary medical specialists, informaticists, and industry participants. These efforts have led toward the development of an Integrating the Healthcare Enterprise (IHE) profile that will serve as a foundation for interoperable interactive multimedia reporting.

Multispecialty Enterprise Imaging Workgroup Consensus on Interactive Multimedia Reporting Current State and Road to the Future: HIMSS-SIIM Collaborative White Paper.

Diagnostic and evidential static image, video clip, and sound multimedia are captured during routine clinical care in cardiology, dermatology, ophthalmology, pathology, physiatry, radiation oncology, radiology, endoscopic procedural specialties, and other medical disciplines. Providers typically describe the multimedia findings in contemporaneous electronic health record clinical notes or associate a textual interpretative report. Visual communication aids commonly used to connect, synthesize, and supplement multimedia and descriptive text outside medicine remain technically challenging to integrate into patient care. Such beneficial interactive elements may include hyperlinks between text, multimedia elements, alphanumeric and geometric annotations, tables, graphs, timelines, diagrams, anatomic maps, and hyperlinks to external educational references that patients or provider consumers may find valuable. This HIMSS-SIIM Enterprise Imaging Community workgroup white paper outlines the current and desired clinical future state of interactive multimedia reporting (IMR). The workgroup adopted a consensus definition of IMR as "interactive medical documentation that combines clinical images, videos, sound, imaging metadata, and/or image annotations with text, typographic emphases, tables, graphs, event timelines, anatomic maps, hyperlinks, and/or educational resources to optimize communication between medical professionals, and between medical professionals and their patients." This white paper also serves as a precursor for future efforts toward solving technical issues impeding routine interactive multimedia report creation and ingestion into electronic health records.

Clinical Application of Image Analysis in Pathology.

Quantitative biomarkers are key prognostic and predictive factors in the diagnosis and treatment of cancer. In the clinical laboratory, the majority of biomarker quantitation is still performed manually, but digital image analysis (DIA) methods have been steadily growing and account for around 25% of all quantitative immunohistochemistry (IHC) testing performed today. Quantitative DIA is primarily employed in the analysis of breast cancer IHC biomarkers, including estrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2/neu; more recently clinical applications have expanded to include human epidermal growth factor receptor 2/neu in gastroesophageal adenocarcinomas and Ki-67 in both breast cancer and gastrointestinal and pancreatic neuroendocrine tumors. Evidence in the literature suggests that DIA has significant benefits over manual quantitation of IHC biomarkers, such as increased objectivity, accuracy, and reproducibility. Despite this fact, a number of barriers to the adoption of DIA in the clinical laboratory persist. These include difficulties in integrating DIA into clinical workflows, lack of standards for integrating DIA software with laboratory information systems and digital pathology systems, costs of implementing DIA, inadequate reimbursement relative to those costs, and other factors. These barriers to adoption may be overcome with international standards such as Digital Imaging and Communications in Medicine (DICOM), increased adoption of routine digital pathology workflows, the application of artificial intelligence to DIA, and the emergence of new clinical applications for DIA.

PD-1 Blockade in Tumors with Mismatch-Repair Deficiency.

BACKGROUND: Somatic mutations have the potential to encode "non-self" immunogenic antigens. We hypothesized that tumors with a large number of somatic mutations due to mismatch-repair defects may be susceptible to immune checkpoint blockade. METHODS: We conducted a phase 2 study to evaluate the clinical activity of pembrolizumab, an anti-programmed death 1 immune checkpoint inhibitor, in 41 patients with progressive metastatic carcinoma with or without mismatch-repair deficiency. Pembrolizumab was administered intravenously at a dose of 10 mg per kilogram of body weight every 14 days in patients with mismatch repair-deficient colorectal cancers, patients with mismatch repair-proficient colorectal cancers, and patients with mismatch repair-deficient cancers that were not colorectal. The coprimary end points were the immune-related objective response rate and the 20-week immune-related progression-free survival rate. RESULTS: The immune-related objective response rate and immune-related progression-free survival rate were 40% (4 of 10 patients) and 78% (7 of 9 patients), respectively, for mismatch repair-deficient colorectal cancers and 0% (0 of 18 patients) and 11% (2 of 18 patients) for mismatch repair-proficient colorectal cancers. The median progression-free survival and overall survival were not reached in the cohort with mismatch repair-deficient colorectal cancer but were 2.2 and 5.0 months, respectively, in the cohort with mismatch repair-proficient colorectal cancer (hazard ratio for disease progression or death, 0.10 [P<0.001], and hazard ratio for death, 0.22 [P=0.05]). Patients with mismatch repair-deficient noncolorectal cancer had responses similar to those of patients with mismatch repair-deficient colorectal cancer (immune-related objective response rate, 71% [5 of 7 patients]; immune-related progression-free survival rate, 67% [4 of 6 patients]). Whole-exome sequencing revealed a mean of 1782 somatic mutations per tumor in mismatch repair-deficient tumors, as compared with 73 in mismatch repair-proficient tumors (P=0.007), and high somatic mutation loads were associated with prolonged progression-free survival (P=0.02). CONCLUSIONS: This study showed that mismatch-repair status predicted clinical benefit of immune checkpoint blockade with pembrolizumab. (Funded by Johns Hopkins University and others; ClinicalTrials.gov number, NCT01876511.).

Low Intratumoral Mast Cells Are Associated With a Higher Risk of Prostate Cancer Recurrence.

BACKGROUND: Mast cells are of interest in prostate cancer because they possess both pro- and anti-tumorigenic properties and are present in the tumor microenvironment. We studied the association of mast cell count and densities with prostate cancer recurrence using tissue microarrays (TMAs) for 462 men who recurred (cases) and 462 controls that were matched to the cases nested in a cohort of radical prostatectomy patients. METHODS: Dual-immunostaining for mast cell tryptase and epithelial cytokeratin-8 and whole slide image analysis were used to assess total mast cell number, mast cell density (mast cell number/tissue area), and mast cell number per epithelial or stromal area in TMA spots containing tumor (up to 4 per man). We used conditional logistic regression to estimate the odds ratio (OR) and 95% confidence interval of recurrence for the mean, minimum, and maximum mast cell parameters in tumor tissue among each man's TMA spots. RESULTS: After taking into account matching factors of age, race, Gleason sum, and pathologic stage, higher minimum mast cell density in the tumor (comparing highest to lowest quartiles: OR = 0.58, 95% CI 0.40-0.86; P-trend = 0.004) was associated with a lower risk of recurrence. Patterns for mast cell number and ratio of mast cell number to epithelial or stromal area were similar to those for mast cell density. CONCLUSIONS: Our results suggest that intratumoral mast cells may be protective against prostate cancer recurrence and could potentially serve as a prognostic biomarker after prostatectomy. Prostate 77: 412-424, 2017. © 2016 Wiley Periodicals, Inc.

Utilization of HPASubC for the Identification of Sinusoid-Specific Proteins in the Liver.

Mass spectrometry-based proteomes of human organs and tissues are powerful tools but fail to capture protein localization and expression at the cellular level. For example, the proteome signal in liver represents the combined protein expression across diverse cellular constituents that include hepatocytes, Kupffer cells, endothelial cells, and others. We utilized HPASubC and the Human Protein Atlas (HPA) to identify the sinusoidal component of protein liver expression to further subset and organize this homogeneous signal. We evaluated 51 109 liver images covering 13 197 proteins from the HPA and discovered 1054 proteins that were exclusive to sinusoidal cells. Sinusoidal staining patterns were identified in a Kupffer cell (n = 247), endothelial cell (n = 358), or lymphocyte (n = 86) specific pattern. Two-hundred and thirty-nine of these proteins were not present in the NextProt or Human Proteome Map liver data sets, potentially expanding our knowledge of the liver proteome. We additionally demonstrate unique endothelial cell expression patterns that distinguish between portal vein, hepatic artery, capillary sinusoids, and central vein regions. These findings significantly improve our understanding of the liver proteome with insight into the endothelial complexity across the hepatic vascular network.

Genome-wide characterization of human L1 antisense promoter-driven transcripts.

BACKGROUND: Long INterspersed Element-1 (LINE-1 or L1) is the only autonomously active, transposable element in the human genome. L1 sequences comprise approximately 17 % of the human genome, but only the evolutionarily recent, human-specific subfamily is retrotransposition competent. The L1 promoter has a bidirectional orientation containing a sense promoter that drives the transcription of two proteins required for retrotransposition and an antisense promoter. The L1 antisense promoter can drive transcription of chimeric transcripts: 5' L1 antisense sequences spliced to the exons of neighboring genes. RESULTS: The impact of L1 antisense promoter activity on cellular transcriptomes is poorly understood. To investigate this, we analyzed GenBank ESTs for messenger RNAs that initiate in the L1 antisense promoter. We identified 988 putative L1 antisense chimeric transcripts, 911 of which have not been previously reported. These appear to be alternative genic transcripts, sense-oriented with respect to gene and initiating near, but typically downstream of, the gene transcriptional start site. In multiple cell lines, L1 antisense promoters display enrichment for YY1 transcription factor and histone modifications associated with active promoters. Global run-on sequencing data support the activity of the L1 antisense promoter. We independently detected 124 L1 antisense chimeric transcripts using long read Pacific Biosciences RNA-seq data. Furthermore, we validated four chimeric transcripts by quantitative RT-PCR and Sanger sequencing and demonstrated that they are readily detectable in many normal human tissues. CONCLUSIONS: We present a comprehensive characterization of human L1 antisense promoter-driven transcripts and provide substantial evidence that they are transcribed in a variety of human cell-types. Our findings reveal a new wide-reaching aspect of L1 biology by identifying antisense transcripts affecting as many as 4 % of all human genes.

Lessons from miR-143/145: the importance of cell-type localization of miRNAs.

miR-143 and miR-145 are co-expressed microRNAs (miRNAs) that have been extensively studied as potential tumor suppressors. These miRNAs are highly expressed in the colon and are consistently reported as being downregulated in colorectal and other cancers. Through regulation of multiple targets, they elicit potent effects on cancer cell growth and tumorigenesis. Importantly, a recent discovery demonstrates that miR-143 and miR-145 are not expressed in colonic epithelial cells; rather, these two miRNAs are highly expressed in mesenchymal cells such as fibroblasts and smooth muscle cells. The expression patterns of miR-143 and miR-145 and other miRNAs were initially determined from tissue level data without consideration that multiple different cell types, each with their own unique miRNA expression patterns, make up each tissue. Herein, we discuss the early reports on the identification of dysregulated miR-143 and miR-145 expression in colorectal cancer and how lack of consideration of cellular composition of normal tissue led to the misconception that these miRNAs are downregulated in cancer. We evaluate mechanistic data from miR-143/145 studies in context of their cell type-restricted expression pattern and the potential of these miRNAs to be considered tumor suppressors. Further, we examine other examples of miRNAs being investigated in inappropriate cell types modulating pathways in a non-biological fashion. Our review highlights the importance of determining the cellular expression pattern of each miRNA, so that downstream studies are conducted in the appropriate cell type.

A single institution's 26-year experience with nonfunctional pancreatic neuroendocrine tumors: a validation of current staging systems and a new prognostic nomogram.

OBJECTIVE: To validate the 2010 American Joint Committee on Cancer (AJCC) and 2006 European Neuroendocrine Tumor Society (ENETS) tumor staging systems for pancreatic neuroendocrine tumors (PanNETs) using the largest, single-institution series of surgically resected patients in the literature. BACKGROUND: The natural history and prognosis of PanNETs have been poorly defined because of the rarity and heterogeneity of these neoplasms. Currently, there are 2 main staging systems for PanNETs, which can complicate comparisons of reports in the literature and thereby hinder progress against this disease. METHODS: Univariate and multivariate analyses were conducted on the prognostic factors of survival using 326 sporadic, nonfunctional, surgically resected PanNET patients who were cared for at our institution between 1984 and 2011. Current and proposed models were tested for survival prognostication validity as measured by discrimination (Harrel's c-index, HCI) and calibration. RESULTS: Five-year overall-survival rates for AJCC stages I, II, and IV are 93% (88%-99%), 74% (65%-83%), and 56% (42%-73%), respectively, whereas ENETS stages I, II, III, and IV are 97% (92%-100%), 87% (80%-95%), 73% (63%-84%), and 56% (42%-73%), respectively. Each model has an HCI of 0.68, and they are no different in their ability to predict survival. We developed a simple prognostic tool just using grade, as measured by continuous Ki-67 labeling, sex, and binary age that has an HCI of 0.74. CONCLUSIONS: Both the AJCC and ENETS staging systems are valid and indistinguishable in their survival prognostication. A new, simpler prognostic tool can be used to predict survival and decrease interinstitutional mistakes and uncertainties regarding these neoplasms.

Human papillomavirus clearance among males is associated with HIV acquisition and increased dendritic cell density in the foreskin.

BACKGROUND: The association between human papillomavirus (HPV) infection and the risk of human immunodeficiency virus (HIV) seroconversion is unclear, and the genital cellular immunology has not been evaluated. METHODS: A case-control analysis nested within a male circumcision trial was conducted. Cases consisted of 44 male HIV seroconverters, and controls were 787 males who were persistently negative for HIV. The Roche HPV Linear Array Genotype Test detected high-risk HPV (HR-HPV) and low-risk HPV (LR-HPV) genotypes. Generalized estimating equations logistic regression was used to estimate adjusted odds ratios (aORs) of HIV seroconversion. In addition, densities of CD1a(+) dendritic cells, CD4(+) T cells, and CD8(+) T cells were measured using immunohistochemistry analysis in foreskins of 79 males randomly selected from participants in the circumcision trial. RESULTS: HR-HPV or LR-HPV acquisition was not significantly associated with HIV seroconversion, after adjustment for sexual behaviors. However, HR-HPV and LR-HPV clearance was significantly associated with HIV seroconversion (aOR, 3.25 [95% confidence interval {CI}, 1.11-9.55] and 3.18 [95% CI, 1.14-8.90], respectively). The odds of HIV seroconversion increased with increasing number of HPV genotypes cleared (P < .001, by the test for trend). The median CD1a(+) dendritic cell density in the foreskin epidermis was significantly higher among males who cleared HPV (72.0 cells/mm(2) [interquartile range {IQR}, 29.4-138.3 cells/mm(2)]), compared with males who were persistently negative for HPV (32.1 cells/mm(2) [IQR, 3.1-96.2 cells/mm(2)]; P = .047), and increased progressively with the number of HPV genotypes cleared (P = .05). CONCLUSIONS: HPV clearance was associated with subsequent HIV seroconversion and also with increased epidermal dendritic cell density, which potentially mediates HIV seroconversion.

Learning with limited target data to detect cells in cross-modality images.

Deep neural networks have achieved excellent cell or nucleus quantification performance in microscopy images, but they often suffer from performance degradation when applied to cross-modality imaging data. Unsupervised domain adaptation (UDA) based on generative adversarial networks (GANs) has recently improved the performance of cross-modality medical image quantification. However, current GAN-based UDA methods typically require abundant target data for model training, which is often very expensive or even impossible to obtain for real applications. In this paper, we study a more realistic yet challenging UDA situation, where (unlabeled) target training data is limited and previous work seldom delves into cell identification. We first enhance a dual GAN with task-specific modeling, which provides additional supervision signals to assist with generator learning. We explore both single-directional and bidirectional task-augmented GANs for domain adaptation. Then, we further improve the GAN by introducing a differentiable, stochastic data augmentation module to explicitly reduce discriminator overfitting. We examine source-, target-, and dual-domain data augmentation for GAN enhancement, as well as joint task and data augmentation in a unified GAN-based UDA framework. We evaluate the framework for cell detection on multiple public and in-house microscopy image datasets, which are acquired with different imaging modalities, staining protocols and/or tissue preparations. The experiments demonstrate that our method significantly boosts performance when compared with the reference baseline, and it is superior to or on par with fully supervised models that are trained with real target annotations. In addition, our method outperforms recent state-of-the-art UDA approaches by a large margin on different datasets.

Cybersecurity and Information Assurance for the Clinical Laboratory.

BACKGROUND: Network-connected medical devices have rapidly proliferated in the wake of recent global catalysts, leaving clinical laboratories and healthcare organizations vulnerable to malicious actors seeking to ransom sensitive healthcare information. As organizations become increasingly dependent on integrated systems and data-driven patient care operations, a sudden cyberattack and the associated downtime can have a devastating impact on patient care and the institution as a whole. Cybersecurity, information security, and information assurance principles are, therefore, vital for clinical laboratories to fully prepare for what has now become inevitable, future cyberattacks. CONTENT: This review aims to provide a basic understanding of cybersecurity, information security, and information assurance principles as they relate to healthcare and the clinical laboratories. Common cybersecurity risks and threats are defined in addition to current proactive and reactive cybersecurity controls. Information assurance strategies are reviewed, including traditional castle-and-moat and zero-trust security models. Finally, ways in which clinical laboratories can prepare for an eventual cyberattack with extended downtime are discussed. SUMMARY: The future of healthcare is intimately tied to technology, interoperability, and data to deliver the highest quality of patient care. Understanding cybersecurity and information assurance is just the first preparative step for clinical laboratories as they ensure the protection of patient data and the continuity of their operations.

Power-law growth models explain incidences and sizes of pancreatic cancer precursor lesions and confirm spatial genomic findings.

Pancreatic ductal adenocarcinoma is a rare but lethal cancer. Recent evidence reveals that pancreatic intraepithelial neoplasms (PanINs), the microscopic precursor lesions in the pancreatic ducts that can give rise to invasive pancreatic cancer, are significantly larger and more prevalent than previously believed. Better understanding of the growth law dynamics of PanINs may improve our ability to understand how a miniscule fraction of these lesions makes the transition to invasive cancer. Here, using artificial intelligence (AI)-based three-dimensional (3D) tissue mapping method, we measured the volumes of >1,000 PanIN and found that lesion size is distributed according to a power law with a fitted exponent of -1.7 over > 3 orders of magnitude. Our data also suggest that PanIN growth is not very sensitive to the pancreatic microenvironment or an individual's age, family history, and lifestyle, and is rather shaped by general growth behavior. We analyze several models of PanIN growth and fit the predicted size distributions to the observed data. The best fitting models suggest that both intraductal spread of PanIN lesions and fusing of multiple lesions into large, highly branched structures drive PanIN growth patterns. This work lays the groundwork for future mathematical modeling efforts integrating PanIN incidence, morphology, genomic, and transcriptomic features to understand pancreas tumorigenesis, and demonstrates the utility of combining experimental measurement of human tissues with dynamic modeling for understanding cancer tumorigenesis.

Three-dimensional genomic mapping of human pancreatic tissue reveals striking multifocality and genetic heterogeneity in precancerous lesions.

Pancreatic intraepithelial neoplasia (PanIN) is a precursor to pancreatic cancer and represents a critical opportunity for cancer interception. However, the number, size, shape, and connectivity of PanINs in human pancreatic tissue samples are largely unknown. In this study, we quantitatively assessed human PanINs using CODA, a novel machine-learning pipeline for 3D image analysis that generates quantifiable models of large pieces of human pancreas with single-cell resolution. Using a cohort of 38 large slabs of grossly normal human pancreas from surgical resection specimens, we identified striking multifocality of PanINs, with a mean burden of 13 spatially separate PanINs per cm3 of sampled tissue. Extrapolating this burden to the entire pancreas suggested a median of approximately 1000 PanINs in an entire pancreas. In order to better understand the clonal relationships within and between PanINs, we developed a pipeline for CODA-guided multi-region genomic analysis of PanINs, including targeted and whole exome sequencing. Multi-region assessment of 37 PanINs from eight additional human pancreatic tissue slabs revealed that almost all PanINs contained hotspot mutations in the oncogene KRAS, but no gene other than KRAS was altered in more than 20% of the analyzed PanINs. PanINs contained a mean of 13 somatic mutations per region when analyzed by whole exome sequencing. The majority of analyzed PanINs originated from independent clonal events, with distinct somatic mutation profiles between PanINs in the same tissue slab. A subset of the analyzed PanINs contained multiple KRAS mutations, suggesting a polyclonal origin even in PanINs that are contiguous by rigorous 3D assessment. This study leverages a novel 3D genomic mapping approach to describe, for the first time, the spatial and genetic multifocality of human PanINs, providing important insights into the initiation and progression of pancreatic neoplasia.

Whole-slide imaging: routine pathologic diagnosis.

Digital pathology systems offer pathologists an alternate, emerging mechanism to manage and interpret information. They offer increasingly fast and scalable hardware platforms for slide scanning and software that facilitates remote viewing, slide conferencing, archiving, and image analysis. Deployed initially and validated largely within the research and biopharmaceutical industries, WSI is increasingly being implemented for direct patient care. Improvements in image quality, scan times, and imageviewing browsers will hopefully allow pathologists to more seamlessly convert to digital pathology, much like our radiology colleagues have done before us. However, WSI creates both opportunities and challenges. Although niche applications of WSI technology for clinical, educational, and research purposes are clearly successful, it is evident that several areas still require attention and careful consideration before more widespread clinical adoption of WSI takes place. These include regulatory issues, development of standards of practice and validation guidelines, workflow modifications, as well as defining situations where WSI technology will really improve practice in a cost-effective way. Current progress on these and other issues, along with improving technology, will no doubt pave the way for increased adoption over the next decade, allowing the pathology community as a whole to harness the true potential of WSI for patient care. The digital decade will likely redefine how pathology is practiced and the role of the pathologist.

Low-Resource Adversarial Domain Adaptation for Cross-Modality Nucleus Detection.

Due to domain shifts, deep cell/nucleus detection models trained on one microscopy image dataset might not be applicable to other datasets acquired with different imaging modalities. Unsupervised domain adaptation (UDA) based on generative adversarial networks (GANs) has recently been exploited to close domain gaps and has achieved excellent nucleus detection performance. However, current GAN-based UDA model training often requires a large amount of unannotated target data, which may be prohibitively expensive to obtain in real practice. Additionally, these methods have significant performance degradation when using limited target training data. In this paper, we study a more realistic yet challenging UDA scenario, where (unannotated) target training data is very scarce, a low-resource case rarely explored for nucleus detection in previous work. Specifically, we augment a dual GAN network by leveraging a task-specific model to supplement the target-domain discriminator and facilitate generator learning with limited data. The task model is constrained by cross-domain prediction consistency to encourage semantic content preservation for image-to-image translation. Next, we incorporate a stochastic, differentiable data augmentation module into the task-augmented GAN network to further improve model training by alleviating discriminator overfitting. This data augmentation module is a plug-and-play component, requiring no modification of network architectures or loss functions. We evaluate the proposed low-resource UDA method for nucleus detection on multiple public cross-modality microscopy image datasets. With a single training image in the target domain, our method significantly outperforms recent state-of-the-art UDA approaches and delivers very competitive or superior performance over fully supervised models trained with real labeled target data.