High risk features in colorectal adenomatous polyps: A multi-institutional study.

High risk features in colorectal adenomatous polyps include size >1 cm and advanced histology: high-grade dysplasia and villous architecture. We investigated whether the diagnostic rates of advanced histology in colorectal adenomatous polyps were similar among institutions across the United States, and if not, could differences be explained by patient age, polyp size, and/or CRC rate. Nine academic institutions contributed data from three pathologists who had signed out at least 100 colorectal adenomatous polyps each from 2018 to 2019 taken from patients undergoing screening colonoscopy. For each case, we recorded patient age and sex, polyp size and location, concurrent CRC, and presence or absence of HGD and villous features. A total of 2700 polyps from 1886 patients (mean age: 61 years) were collected. One hundred twenty-four (5 %) of the 2700 polyps had advanced histology, including 35 (1 %) with HGD and 101 (4 %) with villous features. The diagnostic rate of advanced histology varied by institution from 1.7 % to 9.3 % (median: 4.3 %, standard deviation [SD]: 2.5 %). The rate of HGD ranged from 0 % to 3.3 % (median: 1 %, SD: 1.2 %), while the rate of villous architecture varied from 1 % to 8 % (median: 3.7 %, SD: 2.5 %). In a multivariate analysis, the factor most strongly associated with advanced histology was polyp size >1 cm with an odds ratio (OR) of 31.82 (95 % confidence interval [CI]: 20.52-50.25, p < 0.05). Inter-institutional differences in the rate of polyps >1 cm likely explain some of the diagnostic variance, but pathologic subjectivity may be another contributing factor.

Prognostic implications of tumor histology and microenvironment in surgically resected intrahepatic cholangiocarcinoma: a single institutional experience.

Intrahepatic cholangiocarcinoma (ICC) is a highly aggressive malignant neoplasm. Certain histologic features and the tumor microenvironment may impact disease progression. We aim to characterize the clinicopathologic features of ICC to identify prognostic factors. A total of 50 surgically resected ICC (partial or transplant) cases were analyzed. The cohort included 26 men and 24 women with a median age of 62 years. Eighteen (36%) cases were multifocal ICC with a mean largest tumor size of 6.5 cm. Neoadjuvant and adjuvant chemotherapy was done in eight (16%) and 33 (66%) patients, respectively. Histologically, 42 (84%) were small duct type, seven (14%) large duct type, and one mixed (2%). Thirty (60%) cases showed lymphovascular invasion (LVI) and 11 (22%) with perineural invasion (PNI). Twenty-eight (56%) cases demonstrated dense intratumoral hyaline fibrosis and 18 (36%) with tumor necrosis, each ≥ 10% tumor volume. On follow-up, 35 (70%) patients died of disease after a median disease-specific survival (DSS) of 21 months. Univariate analysis revealed that hyaline fibrosis and adjuvant chemotherapy were associated with better DSS, while tumor size, multifocality, necrosis, and peritumoral neutrophil to lymphocyte ratio were associated with worse DSS. In contrast, age, sex, small vs. large duct types, LVI, and individual inflammatory cell counts were not significant prognostic factors. In summary, ICC is a heterogeneous malignancy with variable clinical courses associated with tumor burden, histology, and microenvironment. Targeting specific components within the tumor microenvironments may be a promising approach for treatment in the future.

Systemic lupus in the era of machine learning medicine.

Artificial intelligence and machine learning applications are emerging as transformative technologies in medicine. With greater access to a diverse range of big datasets, researchers are turning to these powerful techniques for data analysis. Machine learning can reveal patterns and interactions between variables in large and complex datasets more accurately and efficiently than traditional statistical methods. Machine learning approaches open new possibilities for studying SLE, a multifactorial, highly heterogeneous and complex disease. Here, we discuss how machine learning methods are rapidly being integrated into the field of SLE research. Recent reports have focused on building prediction models and/or identifying novel biomarkers using both supervised and unsupervised techniques for understanding disease pathogenesis, early diagnosis and prognosis of disease. In this review, we will provide an overview of machine learning techniques to discuss current gaps, challenges and opportunities for SLE studies. External validation of most prediction models is still needed before clinical adoption. Utilisation of deep learning models, access to alternative sources of health data and increased awareness of the ethics, governance and regulations surrounding the use of artificial intelligence in medicine will help propel this exciting field forward.

SWI/SNF chromatin remodeling complex in pancreatic ductal adenocarcinoma: Clinicopathologic and immunohistochemical study.

The SWItch/Sucrose Non-Fermentable (SWI/SNF) complex is a multimeric protein involved in transcription regulation and DNA damage repair. SWI/SNF complex abnormalities are observed in approximately 14-34 % of pancreatic ductal adenocarcinomas (PDACs). Herein, we evaluated the immunohistochemical expression of a subset of the SWI/SNF complex proteins (ARID1A, SMARCA4/BRG1, SMARCA2/BRM, and SMARCB1/INI1) within our PDAC tissue microarray to determine whether SWI/SNF loss is associated with any clinicopathologic features or patient survival in PDAC. In our cohort, 13 of 353 (3.7 %) PDACs showed deficient SWI/SNF complex expression, which included 11 (3.1 %) with ARID1A loss, 1 (0.3 %) with SMARCA4/BRG1 loss, and 1 (0.3 %) with SMARCA2/BRM loss. All cases were SMARCB1/INI1 proficient. The SWI/SNF-deficient PDACs were more frequently identified in older patients with a mean age of 71.6 years (SD = 7.78) compared to the SWI/SNF-proficient PDACs which occurred at a mean age of 65.2 years (SD = 10.95) (P = 0.013). The SWI/SNF-deficient PDACs were associated with higher histologic grade, compared to the SWI/SNF-proficient PDACs (P = 0.029). No other significant clinicopathologic differences were noted between SWI/SNF-deficient and SWI/SNF-proficient PDACs. On follow-up, no significant differences were seen for overall survival and progression-free survival between SWI/SNF-deficient and SWI/SNF-proficient PDACs (both with P > 0.05). In summary, SWI/SNF-deficient PDACs most frequently demonstrate ARID1A loss. SWI/SNF-deficient PDACs are associated with older age and higher histologic grade. No other significant associations among other clinicopathologic parameters were seen in SWI/SNF-deficient PDACs including survival.

Generating new drug repurposing hypotheses using disease-specific hypergraphs.

The drug development pipeline for a new compound can last 10-20 years and cost over $10 billion. Drug repurposing offers a more time- and cost-effective alternative. Computational approaches based on network graph representations, comprising a mixture of disease nodes and their interactions, have recently yielded new drug repurposing hypotheses, including suitable candidates for COVID-19. However, these interactomes remain aggregate by design and often lack disease specificity. This dilution of information may affect the relevance of drug node embeddings to a particular disease, the resulting drug-disease and drug-drug similarity scores, and therefore our ability to identify new targets or drug synergies. To address this problem, we propose constructing and learning disease-specific hypergraphs in which hyperedges encode biological pathways of various lengths. We use a modified node2vec algorithm to generate pathway embeddings. We evaluate our hypergraph's ability to find repurposing targets for an incurable but prevalent disease, Alzheimer's disease (AD), and compare our ranked-ordered recommendations to those derived from a state-of-the-art knowledge graph, the multiscale interactome. Using our method, we successfully identified 7 promising repurposing candidates for AD that were ranked as unlikely repurposing targets by the multiscale interactome but for which the existing literature provides supporting evidence. Additionally, our drug repositioning suggestions are accompanied by explanations, eliciting plausible biological pathways. In the future, we plan on scaling our proposed method to 800+ diseases, combining single-disease hypergraphs into multi-disease hypergraphs to account for subpopulations with risk factors or encode a given patient's comorbidities to formulate personalized repurposing recommendations.Supplementary materials and code: https://github.com/ayujain04/psb_supplement.

Ameloblastic Carcinoma of the Maxilla: A Rare Case Report and Review of Literature from 1948 to 2021.

Ameloblastic carcinoma is a rare malignant neoplasm arising from the odontogenic epithelium. Ameloblastic carcinoma commonly occurs de novo affecting the posterior segments of the mandible. Presently, only less than 100 cases have been reported arising from the maxilla. We report a unique case of maxillary ameloblastic carcinoma in a 68-year-old male with a 5.6 cm positron emission tomography (PET) avid left maxillary sinus mass. The patient underwent a left maxillectomy which revealed hyperchromatic and pleomorphic tumor cells arranged in a nested and trabecular architecture. The tumor cells showed distinct peripheral palisading with reverse polarization. Areas of bone destruction, necrosis, lymphovascular and perineural invasions, as well as atypical mitoses, were identified. Immunohistochemically, the tumor cells were positive for keratin cocktail (AE1/AE3 and CAM 5.2), keratin 19, p40, and weakly positive for MDM2, while negative for calretinin. Molecular analysis revealed wild-type BRAF; however, alterations in CDKN2A/B, MTAP, RB1, SMARCA4, STK11, FGF12, SETD2, and TP53 were present. This histopathologic and molecular profile supported the diagnosis of ameloblastic carcinoma. There has been no evidence of disease recurrence or metastasis eleven months after the initial diagnosis.

Imaging and Clinical Findings in a Series of Six Cases of Rare Primary Mediastinal Liposarcoma.

Primary mediastinal liposarcoma is a rare, fat-containing malignant lesion that can manifest incidentally with varied imaging appearances. The size and location within the mediastinum can vary among patients. Here, the authors describe the clinical presentation, radiographic characteristics, management, and prognosis in a series of six patients with primary mediastinal liposarcoma. The following case series suggests that even simple-appearing fatty intrathoracic lesions may lead to the development of malignant imaging features. Keywords: Conventional Radiography, CT, MR Imaging, PET/CT, Soft Tissues/Skin, Thorax, Mediastinum ©RSNA, 2022.

The Potential For Bias In Machine Learning And Opportunities For Health Insurers To Address It.

As the use of machine learning algorithms in health care continues to expand, there are growing concerns about equity, fairness, and bias in the ways in which machine learning models are developed and used in clinical and business decisions. We present a guide to the data ecosystem used by health insurers to highlight where bias can arise along machine learning pipelines. We suggest mechanisms for identifying and dealing with bias and discuss challenges and opportunities to increase fairness through analytics in the health insurance industry.

Diffuse Pagetoid Squamous Cell Carcinoma in Situ of the Esophagus: A Rare Case Report and Review of Literature.

Squamous cell carcinoma in situ (SCCIS) with diffuse pagetoid features has been well-described in skin and external genitalia. Diffuse pagetoid SCCIS of the esophagus is extremely rare with only two cases published in the English literature. In this article, we report a rare case of diffuse pagetoid SCCIS of the esophagus in an 89-year-old female with no significant past medical history who presented with dysphagia. Endoscopic examination of the esophagus was remarkable for multiple clean base ulcers spanning 4 cm in the proximal esophagus. Biopsy showed enlarged and hyperchromatic dysplastic cells in the basal half of the epithelium with scattered large individual pagetoid cells as well as several apoptotic dyskeratotic cells in the superficial half of the epithelium. Immunohistochemically, the dysplastic cells were positive for CK7 and p40 with overexpression of p53, and were negative for cytokeratin 20, SOX10, GATA3, CDX2, TTF1. Kreyberg stain was negative for mucin. The histologic features and immunohistochemical profile supported the diagnosis of esophageal diffuse pagetoid SCCIS.

Underdiagnosis bias of artificial intelligence algorithms applied to chest radiographs in under-served patient populations.

Artificial intelligence (AI) systems have increasingly achieved expert-level performance in medical imaging applications. However, there is growing concern that such AI systems may reflect and amplify human bias, and reduce the quality of their performance in historically under-served populations such as female patients, Black patients, or patients of low socioeconomic status. Such biases are especially troubling in the context of underdiagnosis, whereby the AI algorithm would inaccurately label an individual with a disease as healthy, potentially delaying access to care. Here, we examine algorithmic underdiagnosis in chest X-ray pathology classification across three large chest X-ray datasets, as well as one multi-source dataset. We find that classifiers produced using state-of-the-art computer vision techniques consistently and selectively underdiagnosed under-served patient populations and that the underdiagnosis rate was higher for intersectional under-served subpopulations, for example, Hispanic female patients. Deployment of AI systems using medical imaging for disease diagnosis with such biases risks exacerbation of existing care biases and can potentially lead to unequal access to medical treatment, thereby raising ethical concerns for the use of these models in the clinic.

Probabilistic Machine Learning for Healthcare.

Machine learning can be used to make sense of healthcare data. Probabilistic machine learning models help provide a complete picture of observed data in healthcare. In this review, we examine how probabilistic machine learning can advance healthcare. We consider challenges in the predictive model building pipeline where probabilistic models can be beneficial, including calibration and missing data. Beyond predictive models, we also investigate the utility of probabilistic machine learning models in phenotyping, in generative models for clinical use cases, and in reinforcement learning.

Ethical Machine Learning in Healthcare.

The use of machine learning (ML) in healthcare raises numerous ethical concerns, especially as models can amplify existing health inequities. Here, we outline ethical considerations for equitable ML in the advancement of healthcare. Specifically, we frame ethics of ML in healthcare through the lens of social justice. We describe ongoing efforts and outline challenges in a proposed pipeline of ethical ML in health, ranging from problem selection to postdeployment considerations. We close by summarizing recommendations to address these challenges.

Congenital hepatic fibrosis and its mimics: a clinicopathologic study of 19 cases at a single institution.

BACKGROUND: Congenital hepatic fibrosis (CHF) is a rare inherited form of ductal plate malformation associated with polycystic kidney disease. The diagnosis requires histopathologic confirmation, but can be challenging to distinguish from other undefined fibrocystic liver diseases. We aimed to describe the clinicopathologic features of congenital hepatic fibrosis (CHF), with comparisons to other entities that may clinically and/or histologically mimic CHF. METHODS: Nineteen cases that carried a clinical and/or histologic impression of CHF were identified at our institution, of which the histology was reassessed and reappraised into two categories: CHF (n=13) and mimics (n=6). The clinicopathologic features between the two groups were analyzed and compared. RESULTS: The CHF group was further sub-classified into those with clinical suspicion (CHF-c, n=8) and those as incidental histology findings (CHF-i, n=5). Patients of CHF-i were much older than CHF-c or mimics (P<0.05). Male and female were equally affected. Six of 8 CHF-c (66.7%) had concurrent kidney diseases, including 5 polycystic kidney diseases. Five of 6 mimics (83.3%) had various kidney diseases, including nephronophthisis, Alport syndrome, renal agenesis, and nephrolithiasis. None of the CHF-i patients had kidney disease, but 3 were associated with hepatic carcinomas. Histology analysis demonstrated characteristic triads (bile duct abnormalities, portal vein hypoplasia, and fibrosis) in all CHF cases. One mimic had paucity of intrahepatic bile ducts, while the other 5 mimics showed abnormal portal veins and nodular regenerative hyperplasia consistent with hepatoportal sclerosis (HPS). CONCLUSIONS: Our study demonstrates classic histology triad of CHF despite a wide spectrum of clinical presentations. HPS is unexpectedly a clinical mimicker of CHF, which can be distinguished histologically.

Primary Colonic Medullary Carcinoma With Exclusive Squamous Differentiation.

BACKGROUND/AIM: Medullary carcinoma (MC) of the colon is a rare subtype of colorectal adenocarcinoma (CRC) with unique histomorphology and frequent mismatch repair (MMR) deficiency. MC with exclusive squamous differentiation has not been reported. We report an unusual case of MC with squamous differentiation and tested this differentiation potential in other MMR-deficient CRC cases. CASE REPORT: A 68-year-old woman presented with a large ascending colon mass and biopsy showed squamoid tumor morphology with immunoprofile concerning for squamous cell carcinoma (SCC). She underwent right hemicolectomy. Immunohistochemistry and next-generation sequencing (NGS) were performed for tumor classification. Macroscopically, the tumor was large and locally advanced. It metastasized to the lung without lymph node metastasis. Microscopically, the tumor cells were monotonous with cytological features of both MC and SCC. Immunostains were diffusely positive for p40 and CK5/6, but negative for other lineage markers including CDX2, CK20, and SATB2. The tumor was MMR deficient with loss of MLH1 and PMS2. NGS confirmed BRAF V600E mutation. In comparison, a tissue microarray comprising 64 previously diagnosed MMR deficient CRC was tested for squamous differentiation, and only 1 case showed focal CK5/6 expression, but none was positive for p40. CONCLUSION: MC with exclusive squamous differentiation not only posed significant diagnostic challenges, but also unveiled unrecognized differentiation plasticity in this tumor type.

Intimate Partner Violence and Injury Prediction From Radiology Reports.

Intimate partner violence (IPV) is an urgent, prevalent, and under-detected public health issue. We present machine learning models to assess patients for IPV and injury. We train the predictive algorithms on radiology reports with 1) IPV labels based on entry to a violence prevention program and 2) injury labels provided by emergency radiology fellowship-trained physicians. Our dataset includes 34,642 radiology reports and 1479 patients of IPV victims and control patients. Our best model predicts IPV a median of 3.08 years before violence prevention program entry with a sensitivity of 64% and a specificity of 95%. We conduct error analysis to determine for which patients our model has especially high or low performance and discuss next steps for a deployed clinical risk model.

CheXclusion: Fairness gaps in deep chest X-ray classifiers.

Machine learning systems have received much attention recently for their ability to achieve expert-level performance on clinical tasks, particularly in medical imaging. Here, we examine the extent to which state-of-the-art deep learning classifiers trained to yield diagnostic labels from X-ray images are biased with respect to protected attributes. We train convolution neural networks to predict 14 diagnostic labels in 3 prominent public chest X-ray datasets: MIMIC-CXR, Chest-Xray8, CheXpert, as well as a multi-site aggregation of all those datasets. We evaluate the TPR disparity - the difference in true positive rates (TPR) - among different protected attributes such as patient sex, age, race, and insurance type as a proxy for socioeconomic status. We demonstrate that TPR disparities exist in the state-of-the-art classifiers in all datasets, for all clinical tasks, and all subgroups. A multi-source dataset corresponds to the smallest disparities, suggesting one way to reduce bias. We find that TPR disparities are not significantly correlated with a subgroup's proportional disease burden. As clinical models move from papers to products, we encourage clinical decision makers to carefully audit for algorithmic disparities prior to deployment. Our supplementary materials can be found at, http://www.marzyehghassemi.com/chexclusion-supp-3/.

A Review of Challenges and Opportunities in Machine Learning for Health.

Modern electronic health records (EHRs) provide data to answer clinically meaningful questions. The growing data in EHRs makes healthcare ripe for the use of machine learning. However, learning in a clinical setting presents unique challenges that complicate the use of common machine learning methodologies. For example, diseases in EHRs are poorly labeled, conditions can encompass multiple underlying endotypes, and healthy individuals are underrepresented. This article serves as a primer to illuminate these challenges and highlights opportunities for members of the machine learning community to contribute to healthcare.