Utility of Machine Learning to Detect Cytomegalovirus in Digital Hematoxylin and Eosin-Stained Slides.

Rapid and accurate cytomegalovirus (CMV) identification in immunosuppressed or immunocompromised patients presenting with diarrhea is essential for therapeutic management. Due to viral latency, however, the gold standard for CMV diagnosis remains to identify viral cytopathic inclusions on routine hematoxylin and eosin (H&E)-stained tissue sections. Therefore, biopsies may be taken and "rushed" for pathology evaluation. Here, we propose the use of artificial intelligence to detect CMV inclusions on routine H&E-stained whole-slide images to aid pathologists in evaluating these cases. Fifty-eight representative H&E slides from 30 cases with CMV inclusions were identified and scanned. The resulting whole-slide images were manually annotated for CMV inclusions and tiled into 300 × 300 pixel patches. Patches containing annotations were labeled "positive," and these tiles were oversampled with image augmentation to account for class imbalance. The remaining patches were labeled "negative." Data were then divided into training, validation, and holdout sets. Multiple deep learning models were provided with training data, and their performance was analyzed. All tested models showed excellent performance. The highest performance was seen using the EfficientNetV2BO model, which had a test (holdout) accuracy of 99.93%, precision of 100.0%, recall (sensitivity) of 99.85%, and area under the curve of 0.9998. Of 518,941 images in the holdout set, there were only 346 false negatives and 2 false positives. This shows proof of concept for the use of digital tools to assist pathologists in screening "rush" biopsies for CMV infection. Given the high precision, cases screened as "positive" can be quickly confirmed by a pathologist, reducing missed CMV inclusions and improving the confidence of preliminary results. Additionally, this may reduce the need for immunohistochemistry in limited tissue samples, reducing associated costs and turnaround time.

Metallothionein immunohistochemistry has high sensitivity and specificity for detection of Wilson disease.

Diagnosis of Wilson disease (WD) can be difficult because of its protean clinical presentations, but early diagnosis is important because effective treatment is available and can prevent disease progression. Similarly, diagnosis of WD on liver biopsy specimens is difficult due to the wide range of histologic appearances. A stain that could help identify WD patients would be of great value. The goal of this study was to use mass spectrometry-based proteomics to identify potential proteins that are differentially expressed in WD compared to controls, and could serve as potential immunohistochemical markers for screening. Several proteins were differentially expressed in WD and immunohistochemical stains for two (metallothionein (MT) and cytochrome C oxidase copper chaperone (COX17)) were tested and compared to other methods of diagnosis in WD including copper staining and quantitative copper assays. We found diffuse metallothionein immunoreactivity in all liver specimens from patients with WD (n = 20); the intensity of the staining was moderate to strong. This staining pattern was distinct from that seen in specimens from the control groups (none of which showed strong, diffuse staining), which included diseases that may be in the clinical or histologic differential of WD (steatohepatitis (n = 51), chronic viral hepatitis (n = 40), autoimmune hepatitis (n = 50), chronic biliary tract disease (n = 42), and normal liver (n = 20)). COX17 immunostain showed no significant difference in expression between the WD and control groups. MT had higher sensitivity than rhodanine for diagnosis of WD. While the quantitative liver copper assays also had high sensitivity, they require more tissue, have a higher cost, longer turnaround time, and are less widely available than an immunohistochemical stain. We conclude that MT IHC is a sensitive immunohistochemical stain for the diagnosis of WD that could be widely deployed as a screening tool for liver biopsies in which WD is in the clinical or histologic differential diagnosis.

Challenges in the Development, Deployment, and Regulation of Artificial Intelligence in Anatomic Pathology.

Significant advances in artificial intelligence (AI), deep learning, and other machine-learning approaches have been made in recent years, with applications found in almost every industry, including health care. AI is capable of completing a spectrum of mundane to complex medically oriented tasks previously performed only by boarded physicians, most recently assisting with the detection of cancers difficult to find on histopathology slides. Although computers will likely not replace pathologists any time soon, properly designed AI-based tools hold great potential for increasing workflow efficiency and diagnostic accuracy in pathology. Recent trends, such as data augmentation, crowdsourcing for generating annotated data sets, and unsupervised learning with molecular and/or clinical outcomes versus human diagnoses as a source of ground truth, are eliminating the direct role of pathologists in algorithm development. Proper integration of AI-based systems into anatomic-pathology practice will necessarily require fully digital imaging platforms, an overhaul of legacy information-technology infrastructures, modification of laboratory/pathologist workflows, appropriate reimbursement/cost-offsetting models, and ultimately, the active participation of pathologists to encourage buy-in and oversight. Regulations tailored to the nature and limitations of AI are currently in development and, when instituted, are expected to promote safe and effective use. This review addresses the challenges in AI development, deployment, and regulation to be overcome prior to its widespread adoption in anatomic pathology.

Leveraging artificial intelligence to predict ERG gene fusion status in prostate cancer.

BACKGROUND: TMPRSS2-ERG gene rearrangement, the most common E26 transformation specific (ETS) gene fusion within prostate cancer, is known to contribute to the pathogenesis of this disease and carries diagnostic annotations for prostate cancer patients clinically. The ERG rearrangement status in prostatic adenocarcinoma currently cannot be reliably identified from histologic features on H&E-stained slides alone and hence requires ancillary studies such as immunohistochemistry (IHC), fluorescent in situ hybridization (FISH) or next generation sequencing (NGS) for identification. METHODS: OBJECTIVE: We accordingly sought to develop a deep learning-based algorithm to identify ERG rearrangement status in prostatic adenocarcinoma based on digitized slides of H&E morphology alone. DESIGN: Setting, and Participants: Whole slide images from 392 in-house and TCGA cases were employed and annotated using QuPath. Image patches of 224 × 224 pixel were exported at 10 ×, 20 ×, and 40 × for input into a deep learning model based on MobileNetV2 convolutional neural network architecture pre-trained on ImageNet. A separate model was trained for each magnification. Training and test datasets consisted of 261 cases and 131 cases, respectively. The output of the model included a prediction of ERG-positive (ERG rearranged) or ERG-negative (ERG not rearranged) status for each input patch. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: Various accuracy measurements including area under the curve (AUC) of the receiver operating characteristic (ROC) curves were used to evaluate the deep learning model. RESULTS AND LIMITATIONS: All models showed similar ROC curves with AUC results ranging between 0.82 and 0.85. The sensitivity and specificity of these models were 75.0% and 83.1% (20 × model), respectively. CONCLUSIONS: A deep learning-based model can successfully predict ERG rearrangement status in the majority of prostatic adenocarcinomas utilizing only H&E-stained digital slides. Such an artificial intelligence-based model can eliminate the need for using extra tumor tissue to perform ancillary studies in order to assess for ERG gene rearrangement in prostatic adenocarcinoma.

Transient elastography versus liver biopsy: discordance in evaluations for fibrosis and steatosis from a pathology standpoint.

Vibration-controlled transient elastography (VCTE) is a non-invasive method of evaluating liver fibrosis and steatosis. It can easily be performed in the outpatient setting and has been suggested as an alternative to liver biopsy. However, VCTE and biopsy discrepancies commonly occur. Patient characteristics, procedure performance, and liver features can impact the reliability of VCTE results. We identified 82 patients who received VCTE and biopsy within one month to assess how frequently major discrepancies occur and to determine the role of the liver biopsy in this workup. In our study, 35.4% of patients had a major fibrosis discrepancy, which was defined as advanced fibrosis or cirrhosis by VCTE and no to minimal fibrosis on biopsy. This was significantly associated with increased BMI, and liver features including steatohepatitis, inflammation, congestion, and cholestasis were important contributors to discrepancies. All patients with advanced fibrosis or cirrhosis on liver biopsy were appropriately detected by VCTE (n = 28). Detection of steatosis was less sensitive as 19% (n = 4 of 21) of patients with moderate to severe steatosis on biopsy were missed by VCTE. Liver biopsy has been traditionally performed for diagnosis, but with the emergence of non-invasive tools to evaluate for liver fibrosis and steatosis, biopsies are now additionally being performed to confirm findings from noninvasive procedures. Although VCTE is a highly sensitive tool for liver fibrosis, it is not as specific, and therefore, the liver biopsy remains the gold standard for accurate fibrosis assessment.

Clinical significance of pathologic abnormalities in biopsy samples from the appendiceal orifice.

AIMS: Appendiceal orifice mucosa often appears inflamed endoscopically, even when other colonic segments appear normal. Histological findings in biopsy samples taken from endoscopically abnormal mucosa may simulate a variety of inflammatory colitides. We performed this study to evaluate the clinical implications of inflammatory changes isolated to the appendiceal orifice. METHODS AND RESULTS: In this double cohort study, biopsy samples from 26 histologically abnormal appendiceal orifices were reviewed. Twenty-five control cases were culled from endoscopically normal (n = 11) and abnormal (n = 14) appendiceal orifices that were histologically normal. Histological findings were correlated with presentation, medication history, findings at other colonic sites and clinical outcomes. Study cases displayed active inflammation (n = 12), chronic active inflammation (n = 13) or features simulating collagenous colitis (n = 1). Eighteen patients had biopsies taken from other colonic sites; these revealed benign polyps (n = 10) or displayed active (n = 4) or chronic active (n = 4) inflammation. All patients with findings isolated to the appendiceal orifice were asymptomatic at most recent clinical follow-up. Four of eight (50%) of the patients with inflammation in other biopsy samples were ultimately diagnosed with ulcerative colitis, in keeping with the well-established role of the appendix as a 'skip lesion' in that disorder. Control patients presented for screening colonoscopy (n = 19), iron deficiency anaemia (n = 3) or change in bowel habits (n = 3) and none reported gastrointestinal symptoms upon follow-up, regardless of the endoscopic appearance of the appendiceal orifice. CONCLUSION: Isolated inflammation of the appendiceal orifice mucosa should not be regarded as a feature of evolving inflammatory bowel disease or other types of chronic colitis.

The diagnosis of clinically significant oesophageal Candida infections: a reappraisal of clinicopathological findings.

AIMS: Distinguishing true oesophageal Candida infections from oral contaminants is a common diagnostic issue. Historically, histological features believed to indicate true infection included epithelial invasion by pseudohyphae and intraepithelial neutrophils. Whether or not these features correlate with endoscopic lesions, symptoms and response to therapy has never been tested in a large cohort. The aim of this study was to determine whether specific histological features correlate with clinical and endoscopic findings when Candida is found in oesophageal biopsies. METHODS AND RESULTS: We reviewed 271 biopsies in which Candida was detected. Cases were evaluated for the presence of desquamated epithelial cells, location/type of fungal forms, neutrophils, and ulceration. Medical records were reviewed for clinical history, endoscopic lesions, and response to antifungal therapy. Statistical analysis was used to determine whether any histological features significantly correlated with clinical variables. There were 120 males and 151 females with a mean age of 42 years. Fifty-nine per cent had symptoms referable to the oesophagus, particularly dysphagia (36%). Most (73%) patients had abnormal endoscopic findings, with plaques, ulcers, or macroscopic evidence of oesophagitis. Seventy-one per cent of patients with documented antifungal therapy showed symptomatic improvement. Overall, there was no statistically significant correlation between any histological feature and presenting symptoms, endoscopic findings, or response to therapy. Importantly, the lack of pseudohyphae, demonstrable invasion of intact epithelium or neutrophilic infiltrates did not exclude clinically significant infection. CONCLUSIONS: We conclude that detection of Candida in oesophageal biopsies is always potentially clinically significant. Treatment decisions should be made on the basis of an integration of clinical, endoscopic and histological findings.

Applications of Large Language Models in Pathology.

Large language models (LLMs) are transformer-based neural networks that can provide human-like responses to questions and instructions. LLMs can generate educational material, summarize text, extract structured data from free text, create reports, write programs, and potentially assist in case sign-out. LLMs combined with vision models can assist in interpreting histopathology images. LLMs have immense potential in transforming pathology practice and education, but these models are not infallible, so any artificial intelligence generated content must be verified with reputable sources. Caution must be exercised on how these models are integrated into clinical practice, as these models can produce hallucinations and incorrect results, and an over-reliance on artificial intelligence may lead to de-skilling and automation bias. This review paper provides a brief history of LLMs and highlights several use cases for LLMs in the field of pathology.

Utility of SATB2 and MOC-31 Immunostains to Distinguish Between Poorly Differentiated Rectal Adenocarcinoma and Anal Squamous Cell Carcinoma.

OBJECTIVES: Colorectal adenocarcinoma and squamous cell carcinoma (SCC) can arise in the anorectum and present a significant diagnostic challenge when poorly differentiated. Accurate diagnosis can significantly influence management, as the treatments for these conditions involve distinct neoadjuvant chemoradiotherapy regimens. MOC-31 and SATB2 have been utilized as specific markers of glandular differentiation and colorectal origin, respectively, but studies have shown that they may be positive in squamous cell carcinoma of other sites. This raises the concern that MOC-31 and SATB2 may be positive in squamous cell carcinoma of the anorectum, and overreliance on these stains may be a potential diagnostic pitfall in differentiating rectal poorly differentiated adenocarcinoma (PDA) from anal nonkeratinizing SCC. METHODS: We identified biopsies from 10 rectal PDA and 17 anorectal nonkeratinizing SCC cases and stained them for MOC-31 and SATB2. RESULTS: We found that MOC-31 was highly sensitive, being positive in 10/10 cases of rectal PDA, but not specific, as it was also positive in 11/17 SCC cases. In contrast, SATB2 was both sensitive, with positive staining in 10/10 rectal PDA cases, and specific, with negative staining in 17/17 SCC cases. This includes equivocal staining in 4 of these negative SCC cases. MOC-31 had a sensitivity of 100% and specificity of 35.3%, while SATB2 had a sensitivity of 100% and specificity of 100%. CONCLUSIONS: Unlike squamous mucosa of the head and neck, and esophagus, SCC of the anus does not frequently stain positively for SATB2. These data suggest that SATB2 is a reliable marker in distinguishing rectal PDA from anorectal nonkeratinizing SCC, whereas MOC-31 is commonly positive in SCC of the anus. It is also important to note that equivocal SATB2 staining may be seen in SCC.

Nonspecificity of Immunohistochemistry for Mycobacteria Species Using a Rabbit Polyclonal Antibody.

CONTEXT.­: Recent publications have featured immunohistochemistry (IHC) as a sensitive tool for detecting Mycobacterium tuberculosis and nontuberculous mycobacteria, but performance is limited to cases suspected to have mycobacterial infection. OBJECTIVE.­: To examine cross-reactivity of a polyclonal antimycobacterial antibody with various types of pathogens, tissues, and inflammatory patterns. DESIGN.­: Surgical pathology files during a period of 6 years were searched, and 40 cases representing a variety of pathogens, tissue types, and inflammatory responses were retrieved. Cases were stained with a rabbit polyclonal antimycobacterial antibody (Biocare Medical, Pacheco, California). The cases and associated histochemical stains, culture, and molecular results were reviewed by 3 pathologists. RESULTS.­: All 8 cases of mycobacterial infection previously diagnosed by other methods were positive for mycobacteria by IHC. In addition, multiple bacterial and fungal organisms and 1 case of Leishmania amastigotes were also immunoreactive with the mycobacterial IHC. CONCLUSIONS.­: Although highly sensitive for mycobacteria, the polyclonal antibody shows significant cross-reactivity with other organisms. This is a sensitive but nonspecific stain that can be used as an alternative confirmation method for mycobacteria, but attention should be paid to inflammatory reaction and organism morphology when IHC is positive to avoid misdiagnosis.

Clinical and Pathological Correlation in Concomitant Celiac Disease and Eosinophilic Esophagitis Suggests Separate Etiologies.

Introduction. Recently, an increased risk of celiac disease or eosinophilic esophagitis has been postulated among patients with either of these disorders, prompting some to suggest a common underlying mechanism, whereas others maintain that their co-existence is coincidental. Methods. We compared clinical and pathological features of 29 patients meeting criteria for both celiac disease and eosinophilic esophagitis to 26 celiac disease and 26 eosinophilic esophagitis controls to determine whether any distinguished study patients from controls. Results. Eight (28%) study patients presented with symptoms of both celiac disease and eosinophilic esophagitis, whereas 14 (48%) had celiac disease symptoms only and 5 had (17%) esophageal symptoms only. Study patients had similar autoimmune and atopic conditions seen in both control groups. Histological severity of disease, including Marsh II-III duodenal histology (study specimens: 87%; controls: 89%), mean peak esophageal eosinophil counts (study specimens: 55/400x field; controls: 80/400X field, P = .1), and presence of eosinophil microabscesses, scale crust, and subepithelial fibrosis were also similar to controls. Gluten-free diet resolved celiac disease-related symptoms (19 of 20, 95%) and histology (10 of 12, 83%), but not esophageal symptoms or eosinophilia in most study patients. Conclusion. Patients with concomitant celiac disease and eosinophilic esophagitis lack distinguishing features compared to controls with celiac disease or eosinophilic esophagitis alone. The occurrence of both disorders is likely coincidental in most cases.

Follow-up biopsies in gastrointestinal immune checkpoint inhibitor toxicity may show markedly different inflammatory patterns than initial injury.

Colitis is a common manifestation of immune checkpoint inhibitor (ICI) toxicity and can present with varied histologic patterns of inflammation, some of which have been shown to be associated with specific ICI drug types. Although the histologic features of ICI colitis seen at the time of diagnosis have been described, there have been few reports following these patients over time. We evaluated initial and follow-up biopsies in 30 patients with ICI colitis and found that 37% of patients developed a different pattern of injury on follow-up biopsy compared to the initial biopsy. Patients with a different inflammatory pattern were more likely to have restarted ICI therapy before their follow-up biopsy (64%) compared to those without a change in inflammatory pattern (11%; P < 0.01). The majority of these patients had changed ICI drug types (86%). Additionally, many cases changed to an inflammatory bowel disease (IBD)-like pattern (36%), raising a question of de novo IBD. However, all of our patients with an IBD-like pattern experienced sustained resolution of symptoms without steroids or other immunosuppressive medications following discontinuation of ICI therapy, consistent with a diagnosis of ICI toxicity. Our findings suggest that follow-up biopsies in patients with ICI colitis may show a different histology and that this does not necessarily warrant a change in the histologic diagnosis to another disease.

Artificial intelligence for human gunshot wound classification.

Certain features are helpful in the identification of gunshot entrance and exit wounds, such as the presence of muzzle imprints, peripheral tears, stippling, bone beveling, and wound border irregularity. Some cases are less straightforward and wounds can thus pose challenges to an emergency room doctor or forensic pathologist. In recent years, deep learning has shown promise in various automated medical image classification tasks. This study explores the feasibility of using a deep learning model to classify entry and exit gunshot wounds in digital color images. A collection of 2418 images of entrance and exit gunshot wounds were procured. Of these, 2028 entrance and 1314 exit wounds were cropped, focusing on the area around each gunshot wound. A ConvNext Tiny deep learning model was trained using the Fastai deep learning library, with a train/validation split ratio of 70/30, until a maximum validation accuracy of 92.6% was achieved. An additional 415 entrance and 293 exit wound images were collected for the test (holdout) set. The model achieved an accuracy of 87.99%, precision of 83.99%, recall of 87.71%, and F1-score 85.81% on the holdout set. Correctly classified were 88.19% of entrance wounds and 87.71% of exit wounds. The results are comparable to what a forensic pathologist can achieve without other morphologic cues. This study represents one of the first applications of artificial intelligence to the field of forensic pathology. This work demonstrates that deep learning models can discern entrance and exit gunshot wounds in digital images with high accuracy.

Clinical Data Do Not Reliably Predict Duodenal Histology at Follow-up in Celiac Disease: A 13 Center Correlative Study.

Validated nonbiopsy methods to assure duodenal mucosal healing in celiac disease are lacking, yet ongoing mucosal injury is associated with anemia, osteoporosis, and lymphoma. Most providers utilize clinical data as surrogates of mucosal status to avoid additional esophagogastroduodenoscopy. The reliability of such surrogates to predict mucosal recovery has been incompletely evaluated. The aim of this study was to rigorously assess patterns of histologic mucosal recovery at follow-up in celiac disease and to correlate findings with clinical data. Gastrointestinal pathologists from 13 centers evaluated initial and follow-up duodenal biopsies from 181 celiac disease patients. Marsh scores and intraepithelial lymphocytes (IELs)/100 enterocytes were assessed blindly. Histology at follow-up was correlated with symptoms, immunoglobulin A anti-tissue transglutaminase titers and gluten-free diet adherence. Fifty-six/181 (31%) patients had persistent villous blunting and 46/181 (25%) patients had just persistently elevated IELs at follow-up, with only 79/181 (44%) patients having complete histologic remission. IEL normalization (82/181; 45%) lagged villous recovery (125/181;69%). In a minority of patients, villous blunting was limited to proximal duodenal biopsies. No correlation was found between Marsh scores and symptoms, normalization of immunoglobulin A anti-tissue transglutaminase serology, or diet adherence. Children showed greater recovery of Marsh score ( P <0.001) and IELs ( P <0.01) than adults. Persistent mucosal injury is common in celiac disease, with discordant villous/IEL normalization. Pathologist awareness of expected findings in celiac disease follow-up biopsies, including their frequent lack of correlation with clinical data, is important for patient management, and has implications for eligibility criteria for therapeutics currently in development.

Computational pathology in 2030: a Delphi study forecasting the role of AI in pathology within the next decade.

BACKGROUND: Artificial intelligence (AI) is rapidly fuelling a fundamental transformation in the practice of pathology. However, clinical integration remains challenging, with no AI algorithms to date in routine adoption within typical anatomic pathology (AP) laboratories. This survey gathered current expert perspectives and expectations regarding the role of AI in AP from those with first-hand computational pathology and AI experience. METHODS: Perspectives were solicited using the Delphi method from 24 subject matter experts between December 2020 and February 2021 regarding the anticipated role of AI in pathology by the year 2030. The study consisted of three consecutive rounds: 1) an open-ended, free response questionnaire generating a list of survey items; 2) a Likert-scale survey scored by experts and analysed for consensus; and 3) a repeat survey of items not reaching consensus to obtain further expert consensus. FINDINGS: Consensus opinions were reached on 141 of 180 survey items (78.3%). Experts agreed that AI would be routinely and impactfully used within AP laboratory and pathologist clinical workflows by 2030. High consensus was reached on 100 items across nine categories encompassing the impact of AI on (1) pathology key performance indicators (KPIs) and (2) the pathology workforce and specific tasks performed by (3) pathologists and (4) AP lab technicians, as well as (5) specific AI applications and their likelihood of routine use by 2030, (6) AI's role in integrated diagnostics, (7) pathology tasks likely to be fully automated using AI, and (8) regulatory/legal and (9) ethical aspects of AI integration in pathology. INTERPRETATION: This systematic consensus study details the expected short-to-mid-term impact of AI on pathology practice. These findings provide timely and relevant information regarding future care delivery in pathology and raise key practical, ethical, and legal challenges that must be addressed prior to AI's successful clinical implementation. FUNDING: No specific funding was provided for this study.

Semiautomated laser capture microdissection of lung adenocarcinoma cytology samples.

OBJECTIVE: In the past decade molecular diagnostics has changed the clinical management of lung adenocarcinoma patients. Molecular diagnostics, however, is largely dependent on the quantity and quality of the tumor DNA that is retrieved from the tissue or cytology samples. Frequently, patients are diagnosed on cytology specimens where the tumor cells are scattered within the cell block, making selecting for tumor enrichment difficult. In the past we have used laser capture microdissection (LCM) to select for pure populations of tumor cells to increase the sensitivity of molecular assays. This study explores several methods for semiautomated computer-guided LCM. STUDY DESIGN: Hematoxylin and eosin- or TTF-1-immunostained slides from a pleural effusion cell block with metastatic lung adenocarcinoma were used for LCM with either AutoScan or a recently described pattern-matching algorithm, spatially invariant vector quantization (SIVQ), to define morphologic predicates (vectors) to select cells of interest. RESULTS: We retrieved pure populations of tumor cells using both algorithm-guided LCM approaches with slight variations in cellular retrievals. Both methods were semiautomated, requiring minimum technical supervision. CONCLUSION: In this study we demonstrate the first semiautomated, computer-guided LCM of a cytology specimen using SIVQ and AutoScan, a first step towards the long-term goal of integrating LCM into the clinical cytology-molecular workflow.

Neural Network Assisted Pathology Case Identification.

BACKGROUND: Traditionally, cases for cohort selection and quality assurance purposes are identified through structured query language (SQL) searches matching specific keywords. Recently, several neural network-based natural language processing (NLP) pipelines have emerged as an accurate alternative/complementary method for case retrieval. METHODS: The diagnosis section of 1000 pathology reports with the terms "colon" and "carcinoma" were retrieved from our laboratory information system through a SQL query. Each of the reports were labeled as either positive or negative, where cases are considered positive if the case was a primary adenocarcinoma of the colon. Negative cases comprised adenocarcinoma from other sites, metastatic adenocarcinomas, benign conditions, rectal cancers, and other cases that do not fit in the primary colonic adenocarcinoma category. The 1000 cases were randomly separated into training, validation, and holdout sets. A convolutional neural network (CNN) model built using Keras (a neural network library) was trained to identify positive cases, and the model was applied to the holdout set to predict the category for each case. RESULTS: The CNN model classified 141 out of 149 primary colonic adenocarcinoma cases, and 43 out of 51 negative cases correctly, achieving an accuracy of 92% and area under the ROC curve (AUC) of 0.957. CONCLUSION: Trained convolutional neural network models by itself, or as an adjunct to keyword and pattern-based text extraction methods may be used to search for pathology cases of interest with high accuracy.

The Histopathologic Features of Sickle Cell Hepatopathy: A Multi-Institutional Study.

OBJECTIVES: Recent data on hepatic histopathology in patients with sickle cell disease (SCD) are lacking. METHODS: A total of 39 liver biopsies from SCD patients from 4 medical institutes were systematically evaluated. RESULTS: The average age of patients was 27 years; 23 were female. The majority of the patients had hemoglobin SS (33), 3 had hemoglobin SC, and 3 sickle cell trait. Elevated liver functional tests and evaluation for cirrhosis were the main indications for biopsy. At the time of biopsy, most had elevated liver transaminases or hepatomegaly. The most common histopathologic abnormalities were Kupffer cell erythrophagocytosis (76.9%), hemosiderosis (74.4%), sinusoidal dilatation (71.8%), and intrasinusoidal sickled red cells (69.3%). Portal inflammation, lobular inflammation, and bile duct injury were mild to minimal and present in a minority of cases. Advanced fibrosis was present in 28.2% of the cases. CONCLUSIONS: The typical histopathologic features seen in patients with SCD include Kupffer cell erythrophagocytosis, hemosiderosis, sinusoidal dilatation, and intrasinusoidal sickled red cells in a pauci-inflammatory or uninflamed background. Necrosis is less common than reported in older literature. Pathologists should be aware that significant portal and lobular inflammation, interface activity, and bile duct injury are unusual and may be suggestive of other etiologies.

A Study of Thyroid Fine Needle Aspiration of Follicular Adenoma in the "Atypia of Undetermined Significance" Bethesda Category Using Digital Image Analysis.

BACKGROUND: Originally designed for computerized image analysis, ThinPrep is underutilized in that role outside gynecological cytology. It can be used to address the inter/intra-observer variability in the evaluation of thyroid fine needle aspiration (TFNA) biopsy and help pathologists to gain additional insight into thyroid cytomorphology. METHODS: We designed and validated a feature engineering and supervised machine learning-based digital image analysis method using ImageJ and Python scikit-learn . The method was trained and validated from 400 low power (100x) and 400 high power (400x) images generated from 40 TFNA cases. RESULT: The area under the curve (AUC) for receiver operating characteristics (ROC) is 0.75 (0.74-0.82) for model based from low-power images and 0.74 (0.69-0.79) for the model based from high-power images. Cytomorphologic features were synthesized using feature engineering and when performed in isolation, they achieved AUC of 0.71 (0.64-0.77) for chromatin, 0.70 (0.64-0.73) for cellularity, 0.65 (0.60-0.69) for cytoarchitecture, 0.57 (0.51-0.61) for nuclear size, and 0.63 (0.57-0.68) for nuclear shape. CONCLUSION: Our study proves that ThinPrep is an excellent preparation method for digital image analysis of thyroid cytomorphology. It can be used to quantitatively harvest morphologic information for diagnostic purpose.

Histologic Variants of Kaposi Sarcoma in the Gastrointestinal Tract: A Contemporary Multi-institutional Clinicopathologic Analysis of 46 Cases.

Kaposi sarcoma (KS) can pose diagnostic challenges in biopsy specimens. Multiple histologic variants of cutaneous KS have been described; however, the histomorphologic spectrum of gastrointestinal (GI) KS has not been systematically studied. This large series comprehensively evaluated 46 cases of KS involving the GI tract and identified 7 histomorphologic variants, some that have not been previously described. Five of them are inconspicuous but have unique morphologic patterns, including lymphangioma/lymphangiectatic-like (n=17), mucosal hemorrhage/telangiectatic-like (n=17), mucosal inflammation-like (n=15), granulation tissue-like (n=13), and mucosal prolapse-like (n=4) variants. These variants can be easily misdiagnosed or misinterpreted on routine examination if KS is not considered, and if the immunohistochemical stain for human herpesvirus-8 is not performed. The other 2 morphologic variants present as spindle cell proliferations and are the GI stromal tumor-like (n=8) and inflammatory myofibroblastic tumor-like (n=2). These variants raise a broad differential diagnosis of spindle cell tumors of the GI tract and could pose diagnostic challenges. In summary, GI KS lesions exhibit variable, often unconventional histomorphologic patterns. KS should be included in the differential diagnosis even if features of conventional KS are not seen, particularly in limited biopsies in immunocompromised patients, such as those with human immunodeficiency virus infection. Although the clinical significance of these morphologic variants is yet to be determined, they are nonetheless important from a diagnostic standpoint. Misdiagnosis and delay in appropriate management can be avoided by recognizing the morphologic diversity of GI KS and appropriately utilizing the human herpesvirus-8 immunohistochemical stain.