Histopathological images are a rich but incompletely explored data type for studying cancer. Manual inspection is time consuming, making it challenging to use for image data mining. Here we show that convolutional neural networks (CNNs) can be systematically applied across cancer types, enabling comparisons to reveal shared spatial behaviors. We develop CNN architectures to analyze 27,815 hematoxylin and eosin scanned images from The Cancer Genome Atlas for tumor/normal, cancer subtype, and mutation classification. Our CNNs are able to classify TCGA pathologist-annotated tumor/normal status of whole slide images (WSIs) in 19 cancer types with consistently high AUCs (0.995 ± 0.008), as well as subtypes with lower but significant accuracy (AUC 0.87 ± 0.1). Remarkably, tumor/normal CNNs trained on one tissue are effective in others (AUC 0.88 ± 0.11), with classifier relationships also recapitulating known adenocarcinoma, carcinoma, and developmental biology. Moreover, classifier comparisons reveal intra-slide spatial similarities, with an average tile-level correlation of 0.45 ± 0.16 between classifier pairs. Breast cancers, bladder cancers, and uterine cancers have spatial patterns that are particularly easy to detect, suggesting these cancers can be canonical types for image analysis. Patterns for TP53 mutations can also be detected, with WSI self- and cross-tissue AUCs ranging from 0.65-0.80. Finally, we comparatively evaluate CNNs on 170 breast and colon cancer images with pathologist-annotated nuclei, finding that both cellular and intercellular regions contribute to CNN accuracy. These results demonstrate the power of CNNs not only for histopathological classification, but also for cross-comparisons to reveal conserved spatial behaviors across tumors.

Histopathological images are a rich but incompletely explored data type for studying cancer. Manual inspection is time consuming, making it challenging to use for image data mining. Here we show that convolutional neural networks (CNNs) can be systematically applied across cancer types, enabling comparisons to reveal shared spatial behaviors. We develop CNN architectures to analyze 27,815 hematoxylin and eosin slides from The Cancer Genome Atlas for tumor/normal, cancer subtype, and mutation classification. Our CNNs are able to classify tumor/normal status of whole slide images (WSIs) in 19 cancer types with consistently high AUCs (0.995±0.008), as well as subtypes with lower but significant accuracy (AUC 0.87±0.1). Remarkably, tumor/normal CNNs trained on one tissue are effective in others (AUC 0.88±0.11), with classifier relationships also recapitulating known adenocarcinoma, carcinoma, and developmental biology. Moreover, classifier comparisons reveal intra-slide spatial similarities, with average tile-level correlation of 0.45±0.16 between classifier pairs. Breast cancers, bladder cancers, and uterine cancers have spatial patterns that are particularly easy to detect, suggesting these cancers can be canonical types for image analysis. Patterns for TP53 mutations can also be detected, with WSI self-and cross-tissue AUCs ranging from 0.65-0.80. Finally, we comparatively evaluate CNNs on 170 breast and colon cancer images with pathologist-annotated nuclei, finding that both cellular and intercellular regions contribute to CNN accuracy. These results demonstrate the power of CNNs not only for histopathological classification, but also for cross-comparisons to reveal conserved spatial biology.

Background and Aims COVID-19 is associated with hepatocellular liver injury of uncertain significance. We aimed to determine whether development of significant liver injury during hospitalization is related to concomitant medications or processes common in COVID-19 (e.g. ischemia, hyperinflammatory or hypercoagulable states) and to determine whether it can result in liver failure and death. Methods 834 consecutive patients hospitalized with COVID-19 were included. Clinical, medication and laboratory data were obtained at admission and throughout hospitalization using an identified database. Significant liver injury was defined as an AST≥ 5X ULN; ischemia was defined as vasopressor use for a minimum of 2 consecutive days; hyper-inflammatory state as hs-CRP ≥100mg/L and hypercoagulability as D-dimer ≥5mg/L, any time during hospitalization. Results 105 (12.6%) patients developed significant liver injury. Compared to those without significant liver injury, ischemia [OR 4.3 (2.5-7.4, p <0.0001)] and tocilizumab use [OR 3.6 (1.9-7.0, p=0.0001)] were independent predictors of significant liver injury. While AST correlated closely with ALT (R=0.89) throughout hospitalization, AST did not correlate with INR (R= 0.10) or with bilirubin (R=0.09). Death during hospitalization occurred in 136 (16.3%) patients. Multivariate logistic regression showed that significant liver injury was not associated with death [OR 1.4 (0.8-2.6, p=0.2)], while ischemic [OR 2.4 (1.4-4.0, p=0.001)] hypercoagulable [OR 1.7 (1.1-2.6, p=0.02)], and hyperinflammatory [OR 1.9 (1.2-3.1, p=0.02)] disease states were significant predictors of death. Conclusions Liver test abnormalities known to be associated with COVID-19 are secondary to other insults, mostly ischemia or drug-induced liver injury, and do not lead to liver insufficiency or death.

Histopathological images are an integral data type for studying cancer. We show pre-trained convolutional neural networks (CNNs) can be systematically applied across cancer types, enabling comparisons to reveal shared spatial behaviors. We develop CNNs with a common architecture trained on 19 cancer types of The Cancer Genome Atlas (TCGA), analyzing 14459 hematoxylin and eosin scanned frozen tissue images. Our CNNs are based on the Inception-V3 network and classify TCGA pathologist-annotated tumor/normal status of whole slide images in all 19 cancer types with consistently high AUCs (0.995±0.008). Remarkably, CNNs trained on one tissue are effective in others (AUC 0.88±0.11), with classifier relationships recapitulating known adenocarcinoma, carcinoma, and developmental biology. Moreover, classifier comparisons reveal intra-slide spatial similarities, with an average tile-level correlation of 0.45±0.16 between classifier pairs on the TCGA test sets. In particular, the TCGA-trained classifiers had average tile-level correlation of 0.52±0.09 and 0.58±0.08 on hold-out TCGA lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) test sets, respectively. These relations are reflected on two external datasets, i.e., LUAD and LUSC whole slide images of Clinical Proteomic Tumor Analysis Consortium. The CNNs trained on TCGA achieved cross-classification AUCs of 0.75±0.12 and 0.73±0.13 on LUAD and LUSC external validation sets, respectively. These CNNs had average tile-level correlations of 0.38±0.09 and 0.39±0.08 on LUAD and LUSC validation sets, respectively. Breast cancers, bladder cancers, and uterine cancers have spatial patterns that are particularly easy to detect, suggesting these cancers can be canonical types for image analysis. This study illustrates pre-trained CNNs can detect tumor features across a wide range of cancers, suggesting presence of pan-cancer tumor features. These shared features allow combining datasets when analyzing small samples to narrow down the parameter search space of CNN models. Citation Format: Javad Noorbakhsh, Saman Farahmand, Ali Foroughi pour, Sandeep Namburi, Dennis Caruana, David Rimm, Mohammad Soltanieh-ha, Kourosh Zarringhalam, Jeffrey H. Chuang. Deep learning identifies conserved pan-cancer tumor features [abstract]. In: Proceedings of the AACR Virtual Special Conference on Artificial Intelligence, Diagnosis, and Imaging; 2021 Jan 13-14. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(5_Suppl):Abstract nr PO-003.

Study Design: Cross-sectional analysis of completed and terminated spine-related clinical trials in the ClinicalTrials.gov registry. Objective: The aim was to quantify completed and terminated spine-related clinical trials, assess reasons for termination, and determine predictors of termination by comparing characteristics of completed and terminated trials. Summary of Background Data: Clinical trials are key to the advancement of products and procedures related to the spine. Unfortunately, trials may be terminated before completion. ClinicalTrials.gov is a registry and results database maintained by the National Library of Medicine that catalogs trial characteristics and tracks overall recruitment status (eg, ongoing, completed, terminated) for each study as well as reasons for termination. Reasons for trial termination have not been specifically evaluated for spine-related clinical trials. Methods: The ClinicalTrials.gov database was queried on July 20, 2021 for all completed and terminated interventional studies registered to date using all available spine-related search terms. Trial characteristics and reason for termination, were abstracted. Univariate and multivariate analyses were performed determine predictors of trial termination. Results: A total of 969 clinical trials were identified and characterized (833 completed, 136 terminated). Insufficient rate of participant accrual was the most frequently reported reason for trial termination, accounting for 33.8% of terminated trials. Multivariate analysis demonstrated increased odds of trial termination for industry-sponsorship [odds ratio (OR)=1.59] relative to sponsorship from local groups, device studies (OR=2.18) relative to investigations of drug or biological product(s), and phase II (OR=3.07) relative to phase III studies (P<0.05 for each). Conclusions: Spine-related clinical trials were found to be terminated 14% of the time, with insufficient accrual being the most common reason for termination. With significant resources put into clinical studies and the need to advance scientific objectives, predictors, and reasons for trial termination should be considered and optimized to increase the completion rate of trials that are initiated.