Background: Cervical cancer is one of the deadliest cancers affecting women globally. Cervical intraepithelial neoplasia (CIN) assessment using histopathological examination of cervical biopsy slides is subject to interobserver variability. Automated processing of digitized histopathology slides has the potential for more accurate classification for CIN grades from normal to increasing grades of pre-malignancy: CIN1, CIN2, and CIN3. Methodology: Cervix disease is generally understood to progress from the bottom (basement membrane) to the top of the epithelium. To model this relationship of disease severity to spatial distribution of abnormalities, we propose a network pipeline, DeepCIN, to analyze high-resolution epithelium images (manually extracted from whole-slide images) hierarchically by focusing on localized vertical regions and fusing this local information for determining Normal/CIN classification. The pipeline contains two classifier networks: (1) a cross-sectional, vertical segment-level sequence generator is trained using weak supervision to generate feature sequences from the vertical segments to preserve the bottom-to-top feature relationships in the epithelium image data and (2) an attention-based fusion network image-level classifier predicting the final CIN grade by merging vertical segment sequences. Results: The model produces the CIN classification results and also determines the vertical segment contributions to CIN grade prediction. Conclusion: Experiments show that DeepCIN achieves pathologist-level CIN classification accuracy.

Background: Prescreening of biopsies has the potential to improve pathologists' workflow. Tools that identify features and display results in a visually thoughtful manner can enhance efficiency, accuracy, and reproducibility. Machine learning for detection of glomeruli ensures comprehensive assessment and registration of four different stains allows for simultaneous navigation and viewing. Methods: Medical renal core biopsies (4 stains each) were digitized using a Leica SCN400 at ×40 and loaded into the Corista Quantum research platform. Glomeruli were manually annotated by pathologists. The tissue on the 4 stains was registered using a combination of keypoint- and intensity-based algorithms, and a 4-panel simultaneous viewing display was created. Using a training cohort, machine learning convolutional neural net (CNN) models were created to identify glomeruli in all stains, and merged into composite fields of views (FOVs). The sensitivity and specificity of glomerulus detection, and FOV area for each detection were calculated. Results: Forty-one biopsies were used for training (28) and same-batch evaluation (6). Seven additional biopsies from a temporally different batch were also evaluated. A variant of AlexNet CNN, used for object recognition, showed the best result for the detection of glomeruli with same-batch and different-batch evaluation: Same-batch sensitivity 92%, “modified” specificity 89%, average FOV size represented 0.8% of the total slide area; different-batch sensitivity 90%, “modified” specificity 98% and average FOV size 1.6% of the total slide area. Conclusions: Glomerulus detection in the best CNN model shows that machine learning algorithms may be accurate for this task. The added benefit of biopsy registration with simultaneous display and navigation allows reviewers to move from one machine-generated FOV to the next in all 4 stains. Together these features could increase both efficiency and accuracy in the review process.

Background: Determining the site of origin for metastatic well-differentiated neuroendocrine tumors (WDNETs) is challenging, and immunohistochemical (IHC) profiles do not always lead to a definitive diagnosis. We sought to determine if a deep-learning convolutional neural network (CNN) could improve upon established IHC profiles in predicting the site of origin in a cohort of WDNETs from the common primary sites. Materials and Methods: Hematoxylin and eosin (H&E)-stained tissue microarrays (TMAs) were created using 215 WDNETs arising from the known primary sites. A CNN trained and tested on 60% (n = 130) and 40% (n = 85) of these cases, respectively. One hundred and seventy-nine cases had TMA tissue remaining for the IHC analysis. These cases were stained with IHC markers pPAX8, CDX2, SATB2, and thyroid transcription factor-1 (markers of pancreas/duodenum, ileum/jejunum/duodenum, colorectum/appendix, and lung WDNET sites of origin, respectively). The CNN diagnosis was deemed correct if it designated a majority or plurality of the tumor area as the known site of origin. The IHC diagnosis was deemed correct if the most specific marker for a particular site of origin met an H-score threshold determined by two pathologists. Results: When all cases were considered, the CNN correctly identified the site of origin at a lower rate compared to IHC (72% vs. 82%, respectively). Of the 85 cases in the CNN test set, 66 had sufficient TMA material for IHC stains, thus 66 cases were available for a direct case-by-case comparison of IHC versus CNN. The CNN correctly identified 70% of these cases, while IHC correctly identified 76%, a finding that was not statistically significant (P = 0.56). Conclusion: A CNN can identify WDNET site of origin at an accuracy rate close to the current gold standard IHC methods.

Background: Gross imaging of surgical specimens is paramount for the accurate gross examination and diagnosis of disease. Optimized imaging workflow can facilitate consistently high-quality gross photographs, especially in high-volume, metropolitan hospitals such as ours. Most commercial medical gross imaging technology provides ergonomically well-designed hardware, remotely operated cameras, intuitive software interfaces, and automation of workflow. However, these solutions are usually cost-prohibitive and require a large sum of capital budget. Materials and Methods: We applied lean techniques such as value stream mapping (VSM) to design a streamlined and error-free workflow for gross imaging process. We implemented a cost-effective technology, UniTwain, combined with high-resolution webcam to achieve the ideal results. Results: We reduced the mean process time from 600 min to 4.0 min (99.3% decrease in duration); the median process time was reduced from 580 min to 3.0 min. The process efficiency increased from 20% to 100%. The implemented solution has a comparable durability, scalability, and archiving feasibility to commercial medical imaging systems and costs four times less. The only limitations are manual operation of the webcam and lower resolution. The webcam sensors have 8.2 megapixel (MP) resolution, approximately 12 MP less than medical imaging devices. However, we believe that this difference is not visually significant and the effect on gross diagnosis with the naked eye is minimal. Conclusions: To our knowledge, this is the first study that utilized UniTwain as a viable, low-cost solution to streamline the gross imaging workflow. The UniTwain combined with high-resolution webcam could be a suitable alternative for our institution that does not plan to heavily invest in medical imaging.

Background: Whole-slide images (WSI) are produced by a high-resolution scanning of pathology glass slides. There are a large number of whole-slide imaging scanners, and the resulting images are frequently larger than 100,000 × 100,000 pixels which typically image 100,000 to one million cells, ranging from several hundred megabytes to many gigabytes in size. Aims and Objectives: Provide HTTP access over the web to Whole Slide Image tiles that do not have localized tiling servers but only basic HTTP access. Move all image decode and tiling functions to calling agent (ImageBox). Methods: Current software systems require tiling image servers to be installed on systems providing local disk access to these images. ImageBox2 breaks this requirement by accessing tiles from remote HTTP source via byte-level HTTP range requests. This method does not require changing the client software as the operation is relegated to the ImageBox2 server which is local (or remote) to the client and can access tiles from remote images that have no server of their own such as Amazon S3 hosted images. That is, it provides a data service [on a server that does not need to be managed], the definition of serverless execution model increasingly favored by cloud computing infrastructure. Conclusions: The specific methodology described and assessed in this report preserves normal client connection semantics by enabling cloud-friendly tiling, promoting a web of http connected whole-slide images from a wide-ranging number of sources, and providing tiling where local tiling servers would have been otherwise unavailable.

Introduction: The initial point in the diagnostic workup of solid tumors remains manual, with the assessment of hematoxylin and eosin (H&E)-stained tissue sections by microscopy. This is a labor-intensive step that requires attention to detail. In addition, diagnoses are influenced by an individual pathologist's knowledge and experience and may not always be reproducible between pathologists. Methods: We introduce a deep learning-based method in colorectal cancer detection and segmentation from digitized H&E-stained histology slides. Results: In this study, we demonstrate that this neural network approach produces median accuracy of 99.9% for normal slides and 94.8% for cancer slides compared to pathologist-based diagnosis on H&E-stained slides digitized from clinical samples. Conclusion: Given that our approach has very high accuracy on normal slides, use of neural network algorithms may provide a screening approach to save pathologist time in identifying tumor regions. We suggest that this new method may be a powerful assistant for colorectal cancer diagnostics.

Background: Whole-slide images (WSIs) as a kind of image data are rapidly growing in the digital pathology domain. With unusual high resolution, these images make them hard to be supported by conventional tools or file formats. Thus, it obstructs data sharing and automated analysis. Here, we propose a library, LibMI, along with its open and standardized image file format. They can be used together to efficiently read, write, modify, and annotate large images. Materials and Methods: LibMI utilizes the concept of pyramid image structure and lazy propagation from a segment tree algorithm to support reading and modifying and to guarantee that both operations have linear time complexity. Further, a cache mechanism was introduced to speed up the program. Results: LibMI is an open and efficient library for histopathological image processing. To demonstrate its functions, we applied it to several tasks including image thresholding, microscopic color correction, and storing pixel-wise information on WSIs. The result shows that libMI is particularly suitable for modifying large images. Furthermore, compared with congeneric libraries and file formats, libMI and modifiable multiscale image (MMSI) run 18.237 times faster on read-only tasks. Conclusions: The combination of libMI library and MMSI file format enables developers to efficiently read and modify WSIs, thus can assist in pixel-wise image processing on extremely large images to promote building image processing pipeline. The library together with the data schema is freely available on GitLab: https://gitlab.com/BioAI/libMI.

Unlocking the full potential of pathology data by gaining computational access to histological pixel data and metadata (digital pathology) is one of the key promises of computational pathology. Despite scientific progress and several regulatory approvals for primary diagnosis using whole-slide imaging, true clinical adoption at scale is slower than anticipated. In the U.S., advances in digital pathology are often siloed pursuits by individual stakeholders, and to our knowledge, there has not been a systematic approach to advance the field through a regulatory science initiative. The Alliance for Digital Pathology ( the Alliance) is a recently established, volunteer, collaborative, regulatory science initiative to standardize digital pathology processes to speed up innovation to patients. The purpose is: (1) to account for the patient perspective by including patient advocacy; (2) to investigate and develop methods and tools for the evaluation of effectiveness, safety, and quality to specify risks and benefits in the precompetitive phase; (3) to help strategize the sequence of clinically meaningful deliverables; (4) to encourage and streamline the development of ground-truth data sets for machine learning model development and validation; and (5) to clarify regulatory pathways by investigating relevant regulatory science questions. The Alliance accepts participation from all stakeholders, and we solicit clinically relevant proposals that will benefit the field at large. The initiative will dissolve once a clinical, interoperable, modularized, integrated solution (from tissue acquisition to diagnostic algorithm) has been implemented. In times of rapidly evolving discoveries, scientific input from subject-matter experts is one essential element to inform regulatory guidance and decision-making. The Alliance aims to establish and promote synergistic regulatory science efforts that will leverage diverse inputs to move digital pathology forward and ultimately improve patient care.

Background: To improve communication between clinical providers and the laboratory, we recently implemented secure text messaging for our critical value notifications. This was done to communicate laboratory critical values (CV) to providers faster so changes to patient care could be done faster. Our previous method of communicating CV to providers was paging and relied on a call back to receive the critical value. Methods: We implemented delivery of CV through a secure texting application in which the CV was directly communicated to the provider on their smart phone device. Results: The mean pre-implementation turnaround time (TAT) was 11.3 minutes (median: 7 minutes, range: 0 - 210 minutes). The mean post- secure text messaging implementation TAT was 3.03 minutes (median: 0.89 minutes, range: < 1 - 95 minutes).When comparing pre- and post-implementation, there was a significant reduction in the TAT from using secure text messaging (p < 0.001). Of the 234 surveys sent out, 81 providers responded (35%). Of these responses, 85% reported that critical value notification by secure text messaging has increased their efficiency and 95% reported that critical value notification is more effective than a pager-phone-call based system. 83% of providers reported that they were able to provide better, faster care to their patients. Conclusions: Using secure text messaging (STM) to deliver critical values significantly reduces the CV TAT. Furthermore, providers noted they preferred to receive CV notifications through STM and reported that they were able to provide more effective care to their patients.

Introduction: In this study, we report on our experience using digital pathology to overcome the severe limitations imposed on health care by the Covid-19 outbreak in Northern Italy. Social distancing had a major impact on public transportation, causing it to run with reduced timetables. This resulted in a major challenge for hospital commuters. To limit the presence in our hospital of no more than two pathologists at a time out of four, a web-based digital pathology system (DPS) was employed to work remotely. Subjects and Methods: We used a DPS in which a scanner, a laboratory information system, a storage device, and a web server were interfaced so that tissue slides could be viewed over the Internet by whole-slide imaging (WSI). After a brief internal verification test, the activity on the DPS was recorded, taking track of a set of performance and efficiency indicators. At the end of the study, 405 cases were signed out remotely. Results: Of 693 cases, 58.4% were signed out remotely by WSI, while 8.4% needed to be kept on hold to return to the original microscope slide. In three cases, at least one slide had to be rescanned. In eight cases, one slide was recut. Panel discussion by WSI was necessary in 34 cases, a condition in which all pathologists were asked for their opinion. A consultation with a more experienced colleague was necessary in 17 cases. Conclusions: We show that WSI easily allows pathologists to overcome the problems caused by the severe social distancing measures imposed by the Covid-19 pandemic. Our experience shows that soon there will not be alternatives to digital pathology, given that there is no assurance that other similar outbreaks will not occur.

Introduction: Teleneuropathology at our institution evolved over the last 17 years from using static to dynamic robotic microscopy. Historically (2003–2007), using older technology, the deferral rate was 19.7%, and the concordance was 81% with the final diagnosis. Two years ago, we switched to use hybrid robotic devices to perform these intraoperative (IO) consultations because our older devices were obsolete. The aim of this study was to evaluate the impact this change had on our deferral and concordance rates with teleneuropathology using this newer instrument. Materials and Methods: Aperio LV1 4-slide capacity hybrid robotic scanners with an attached desktop console (Leica Biosystems, Vista, CA, USA) and GoToAssist (v4.5.0.1620, Boston, MA, USA) were used for IO telepathology cases. A cross-sectional comparative study was conducted comparing teleneuropathology from three remote hospitals (193 cases) to IO neuropathology consultation performed by conventional glass slide examination at a light microscope (310 cases) from the host hospital. Deferral and concordance rates were compared to final histopathological diagnoses. Results: The deferral rate for IO teleneuropathology was 26% and conventional glass slide 24.24% (P = 0.58). The concordance rate for teleneuropathology was 93.94%, which was slightly higher than 89.09% for conventional glass slides (P = 0.047). Conclusion: The new hybrid robotic device for performing IO teleneuropathology interpretations at our institution was as effective as conventional glass slide interpretation. While we did observe a noticeable change in the deferral rate compared to prior years, we did appreciate the marked improvement of the concordance rate using this new hybrid scanner.

Background: Cervical screening could potentially be improved by better stratifying individual risk for the development of cervical cancer or precancer, possibly even allowing follow-up of individual patients differently than proposed under current guidelines that focus primarily on recent screening test results. We explore the use of a Bayesian decision science model to quantitatively stratify individual risk for the development of cervical squamous neoplasia. Materials and Methods: We previously developed a dynamic multivariate Bayesian network model that uses cervical screening and histopathologic data collected over 13 years in our system to quantitatively estimate the risk of individuals for the development of cervical precancer or invasive cervical cancer. The database includes 1,126,048 liquid-based cytology test results belonging to 389,929 women. From-the-vial, high risk human papilloma virus (HPV) test results and follow-up gynecological surgical procedures were available on 33.6% and 12% of these results (378,896 and 134,727), respectively. Results: Historical data impacted 5-year cumulative risk for both histopathologic cervical intraepithelial neoplasia 3 (CIN3) and squamous cell carcinoma (SCC) diagnoses. The risk was highest in patients with prior high grade squamous intraepithelial lesion cytology results. Persistent abnormal cervical screening test results, either cytologic or HPV results, were associated with variable increasing risk for squamous neoplasia. Risk also increased with prior histopathologic diagnoses of precancer, including CIN2, CIN3, and adenocarcinoma in situ. Conclusions: Bayesian modeling allows for individualized quantitative risk assessments of system patients for histopathologic diagnoses of significant cervical squamous neoplasia, including very rare outcomes such as SCC.

Health care is undergoing a profound transformation driven by an increase in new types of diagnostic data, increased data sharing enabled by interoperability, and improvements in our ability to interpret data through the application of artificial intelligence and machine learning. Paradoxically, we are also discovering that our current paradigms for implementing electronic health-care records and our ability to create new models for reforming the health-care system have fallen short of expectations. This article traces these shortcomings to two basic issues. The first is a reliance on highly centralized quality improvement and measurement strategies that fail to account for the high level of variation and complexity found in human disease. The second is a reliance on legacy payment systems that fail to reward the sharing of data and knowledge across the health-care system. To address these issues, and to better harness the advances in health care noted above, the health-care system must undertake a phased set of reforms. First, efforts must focus on improving both the diagnostic process and data sharing at the local level. These efforts should include the formation of diagnostic management teams and increased collaboration between pathologists and radiologists. Next, building off current efforts to develop national federated research databases, providers must be able to query national databases when information is needed to inform the care of a specific complex patient. In addition, providers, when treating a specific complex patient, should be enabled to consult nationally with other providers who have experience with similar patient issues. The goal of these efforts is to build a health-care system that is funded in part by a novel fee-for-knowledge-sharing paradigm that fosters a collaborative decentralized approach to patient care and financially incentivizes large-scale data and knowledge sharing.

Background: Little is known about the effect of a minimum number of slides required in generating image datasets used to build generalizable machine-learning (ML) models. In addition, the assumption within deep learning is that the increased number of training images will always enhance accuracy and that the initial validation accuracy of the models correlates well with their generalizability. In this pilot study, we have been able to test the above assumptions to gain a better understanding of such platforms, especially when data resources are limited. Methods: Using 10 colon histology slides (5 carcinoma and 5 benign), we were able to acquire 1000 partially overlapping images (Dataset A) that were then trained and tested on three convolutional neural networks (CNNs), ResNet50, AlexNet, and SqueezeNet, to build a large number of unique models for a simple task of classifying colon histopathology into benign and malignant. Different quantities of images (10–1000) from Dataset A were used to construct >200 unique CNN models whose performances were individually assessed. The performance of these models was initially assessed using 20% of Dataset A's images (not included in the training phase) to acquire their initial validation accuracy (internal accuracy) followed by their generalization accuracy on Dataset B (a very distinct secondary test set acquired from public domain online sources).Results: All CNNs showed similar peak internal accuracies (>97%) from the Dataset A test set. Peak accuracies for the external novel test set (Dataset B), an assessment of the ability to generalize, showed marked variation (ResNet50: 98%; AlexNet: 92%; and SqueezeNet: 80%). The models with the highest accuracy were not generated using the largest training sets. Further, a model's internal accuracy did not always correlate with its generalization accuracy. The results were obtained using an optimized number of cases and controls. Conclusions: Increasing the number of images in a training set does not always improve model accuracy, and significant numbers of cases may not always be needed for generalization, especially for simple tasks. Different CNNs reach peak accuracy with different training set sizes. Further studies are required to evaluate the above findings in more complex ML models prior to using such ancillary tools in clinical settings.

Background: Free-text sections of pathology reports contain the most important information from a diagnostic standpoint. However, this information is largely underutilized for computer-based analytics. The vast majority of NLP-based methods lack a capacity to accurately extract complex diagnostic entities and relationships among them as well as to provide an adequate knowledge representation for downstream data-mining applications. Methods: In this paper, we introduce a novel informatics pipeline that extends open information extraction (openIE) techniques with artificial intelligence (AI) based modeling to extract and transform complex diagnostic entities and relationships among them into Knowledge Graphs (KGs) of relational triples (RTs). Results: Evaluation studies have demonstrated that the pipeline's output significantly differs from a random process. The semantic similarity with original reports is high (Mean Weighted Overlap of 0.83). The precision and recall of extracted RTs based on experts' assessment were 0.925 and 0.841 respectively (P <0.0001). Inter-rater agreement was significant at 93.6% and inter-rated reliability was 81.8%. Conclusion: The results demonstrated important properties of the pipeline such as high accuracy, minimality and adequate knowledge representation. Therefore, we conclude that the pipeline can be used in various downstream data-mining applications to assist diagnostic medicine.

Background: Previous studies have demonstrated the noninferiority of pathologists' interpretation of whole slide images (WSIs) compared to microscopic slides in diagnostic surgical pathology; however, to our knowledge, no published studies have tested analytical precision of an entire WSI system. Methods: In this study, five pathologists at three locations tested intra-system, inter-system/site, and intra- and inter-pathologist precision of the Aperio AT2 DX System (Leica Biosystems, Vista, CA, USA). Sixty-nine microscopic slides containing 23 different morphologic features suggested by the Digital Pathology Association as important to diagnostic pathology were identified and scanned. Each of 202 unique fields of view (FOVs) had 1–3 defined morphologic features, and each feature was represented in three different tissues. For intra-system precision, each site scanned 23 slides at three different times and one pathologist interpreted all FOVs. For inter-system/site precision, all 69 slides were scanned once at each of three sites, and FOVs from each site were read by one pathologist. To test intra- and inter-pathologist precision, all 69 slides were scanned at one site, FOVs were saved in three different orientations, and the FOVs were transferred to a different site. Three different pathologists then interpreted FOVs from all 69 slides. Wildcard (unscored) slides and washout intervals were included in each study. Agreement estimates with 95% confidence intervals were calculated. Results: Combined precision from all three studies, representing 606 FOVs in each of the three studies, showed overall intra-system agreement of 97.9%; inter-system/site agreement was 96%, intra-pathologist agreement was 95%, and inter-pathologist agreement was 94.2%. Conclusions: Pathologists using the Aperio AT2 DX System identified histopathological features with high precision, providing increased confidence in using WSI for primary diagnosis in surgical pathology.