Standard of care diagnostic procedure for suspected skin cancer is microscopic examination of hematoxylin & eosin stained tissue by a pathologist. Areas of high inter-pathologist discordance and rising biopsy rates necessitate higher efficiency and diagnostic reproducibility. We present and validate a deep learning system which classifies digitized dermatopathology slides into 4 categories. The system is developed using 5,070 images from a single lab, and tested on an uncurated set of 13,537 images from 3 test labs, using whole slide scanners manufactured by 3 different vendors. The system's use of deep-learning-based confidence scoring as a criterion to consider the result as accurate yields an accuracy of up to 98%, and makes it adoptable in a real-world setting. Without confidence scoring, the system achieved an accuracy of 78%. We anticipate that our deep learning system will serve as a foundation enabling faster diagnosis of skin cancer, identification of cases for specialist review, and targeted diagnostic classifications.Every year in the United States, 12 million skin lesions are biopsied 1 , with over 5 million new skin cancer cases diagnosed 2 . After a skin lesion is biopsied, the tissue is fixed, embedded, sectioned, and stained with hematoxylin and eosin (H&E) on glass slides, ultimately to be examined under microscope by a dermatologist, general pathologist or dermatopathologist who provides a diagnosis for each tissue specimen. Owing to the large variety of over 500 distinct skin pathologies 3 and the severe consequences of a critical misdiagnosis 4 , diagnosis in dermatopathology demands specialized training and education. Although the inter-observer concordance rate in dermatopathology is estimated to be between 90 and 95% 5,6 , there are some distinctions which present frequent disagreement among pathologists, such as in the case of melanoma vs. melanocytic nevi 7-11 . Any system which could improve diagnostic accuracy provides obvious benefits for dermatopathology labs and patients; however, there are substantial benefits also to improving the distribution of pathologists' workloads 12-14 . This can reduce diagnostic turnaround times in several scenarios. For example, when skin biopsies are interpreted initially by a dermatologist or a general pathologist, prior to referral to a dermatopathologist, it can result in a delay of days, sometimes in critical cases. In another common scenario, additional staining is required to identify characteristics of the tissue not captured by standard H&E staining. If those additional stains are not ordered early enough, there can be further delays to diagnosis. An intelligent system to distribute pathology workloads could alleviate some of these bottlenecks in lab workflows. The rise in adoption of digital pathology 1,15 provides an opportunity for the use of deep learning-based methods for closing these gaps in diagnostic reliability and efficiency 16,17 .

Background:Artificial intelligence is advancing at an accelerated pace into clinical applications, providing opportunities for increased efficiency, improved accuracy, and cost savings through computer-aided diagnostics. Dermatopathology, with emphasis on pattern recognition, offers a unique opportunity for testing deep learning algorithms.Aims:This study aims to determine the accuracy of deep learning algorithms to diagnose three common dermatopathology diagnoses.Methods:Whole slide images (WSI) of previously diagnosed nodular basal cell carcinomas (BCCs), dermal nevi, and seborrheic keratoses were annotated for areas of distinct morphology. Unannotated WSIs, consisting of five distractor diagnoses of common neoplastic and inflammatory diagnoses, were included in each training set. A proprietary fully convolutional neural network was developed to train algorithms to classify test images as positive or negative relative to ground truth diagnosis.Results:Artificial intelligence system accurately classified 123/124 (99.45%) BCCs (nodular), 113/114 (99.4%) dermal nevi, and 123/123 (100%) seborrheic keratoses.Conclusions:Artificial intelligence using deep learning algorithms is a potential adjunct to diagnosis and may result in improved workflow efficiencies for dermatopathologists and laboratories.

represents a transformative technology that impacts dermatologists and dermatopathologists from residency to academic and private practice. Two concerns are accuracy of interpretation from whole-slide images (WSI) and effect on workflow. Studies of considerably large series involving single-organ systems are lacking.OBJECTIVE To evaluate whether diagnosis from WSI on a digital microscope is inferior to diagnosis of glass slides from traditional microscopy (TM) in a large cohort of dermatopathology cases with attention on image resolution, specifically eosinophils in inflammatory cases and mitotic figures in melanomas, and to measure the workflow efficiency of WSI compared with TM. DESIGN, SETTING, AND PARTICIPANTS Three dermatopathologists established interobserver ground truth consensus (GTC) diagnosis for 499 previously diagnosed cases proportionally representing the spectrum of diagnoses seen in the laboratory. Cases were distributed to 3 different dermatopathologists who diagnosed by WSI and TM with a minimum 30-day washout between methodologies. Intraobserver WSI/TM diagnoses were compared, followed by interobserver comparison with GTC. Concordance, major discrepancies, and minor discrepancies were calculated and analyzed by paired noninferiority testing. We also measured pathologists' read rates to evaluate workflow efficiency between WSI and TM. This retrospective study was caried out in an independent, national, university-affiliated dermatopathology laboratory. MAIN OUTCOMES AND MEASURESIntraobserver concordance of diagnoses between WSI and TM methods and interobserver variance from GTC, following College of American Pathology guidelines.RESULTS Mean intraobserver concordance between WSI and TM was 94%. Mean interobserver concordance was 94% for WSI and GTC and 94% for TM and GTC. Mean interobserver concordance between WSI, TM, and GTC was 91%. Diagnoses from WSI were noninferior to those from TM. Whole-slide image read rates were commensurate with WSI experience, achieving parity with TM by the most experienced user.CONCLUSIONS AND RELEVANCE Diagnosis from WSI was found equivalent to diagnosis from glass slides using TM in this statistically powerful study of 499 dermatopathology cases. This study supports the viability of WSI for primary diagnosis in the clinical setting.

Biopsies were taken from 4 patients who presented to their dermatologist with violaceous papules and plaques of the dorsal toes (COVID Toes) associated with varying degrees of severe acute respiratory syndrome coronavirus 2 exposure and COVID-19 testing. Major histopathologic findings were lymphocytic eccrine inflammation and a spectrum of vasculopathic findings to include superficial and deep angiocentric-perivascular lymphocytic inflammation, lymphocytes in vessel walls (lymphocytic vasculitis), endothelial swelling, red blood cell extravasation, and focal deposits of fibrin in both vessel lumina, and vessel walls. Interface changes were observed to include vacuolopathy and apoptotic keratinocytes at the basement membrane. Immunostains showed a dominant T-cell lineage (positive for T-cell receptor beta, CD2, CD3, CD5, and CD7). B-cells were rare and clusters of CD123positive dermal plasmacytoid dendritic cells were observed surrounding eccrine clusters and some perivascular zones. The consistent perieccrine and vasculopathic features represent important pathologic findings in the diagnosis of COVID toes and are suggestive of pathogenetic mechanisms. Clinicopathologic correlation, the epidemiological backdrop, and the current worldwide COVID-19 pandemic favor a viral causation and should alert the physician to initiate a workup and the appropriate use of COVID-19 testing.

Radioisotopic, pH 6.8 assays were designed to measure hepatic cortisol sulfation in chickens, gerbils, and hamsters of both sexes. Enzyme levels with 40 microM cortisol were similar in males of all three species and due mostly to low Km enzymes with 10-30 microM cortisol Km's. Maximum enzyme activity in male chickens required 40 microM cortisol. In the other species, the much higher maximum enzyme activity observed required 500 microM cortisol owing to sulfotransferases with Km's for the hormone near 300 microM. Coenzyme 3'-phosphoadenosine-5'-phosphosulfate requirements also varied between species. Sex differences of the enzyme levels were found only in hamsters. There, males possessed only 24-33% of the enzyme levels found in females. Cortisol 21-sulfate was the reaction product in all of the species. Sexual dimorphism in hamsters appeared to be due to repressive effects of androgens. pH optima of enzyme activities in the three species ranged from pH 6 to 7. Routine use of pH 6.8 assays allowed representative interspecies comparisons. DEAE-Sephadex fractionation of cytosol showed that chicken liver contained mostly two enzymes with different pH optima that catalyzed cortisol sulfation. These differed from the enzymes that catalyzed dehydroepiandrosterone and estradiol sulfation. In the gerbil four enzymes with similar pH optima catalyzed cortisol sulfation. The second of these to elute from DEAE-Sephadex columns was the low Km form. In hamsters most glucocorticoid sulfotransferase activity appeared to be due to one enzyme. The molecular weights of the low Km gerbil enzyme and the main hamster enzyme were 98 300 +/- 6100 and 105 000 +/- 8100. Hamsters and gerbils responded to injection of cortisol by hepatic tyrosine aminotransferase induction.