A deep learning model and human-machine fusion for prediction of EBV-associated gastric cancer from histopathology

Abstract

Epstein–Barr virus-associated gastric cancer (EBVaGC) shows a robust response to immune checkpoint inhibitors. Therefore, a cost-efficient and accessible tool is needed for discriminating EBV status in patients with gastric cancer. Here we introduce a deep convolutional neural network called EBVNet and its fusion with pathologists for predicting EBVaGC from histopathology. The EBVNet yields an averaged area under the receiver operating curve (AUROC) of 0.969 from the internal cross validation, an AUROC of 0.941 on an external dataset from multiple institutes and an AUROC of 0.895 on The Cancer Genome Atlas dataset. The human-machine fusion significantly improves the diagnostic performance of both the EBVNet and the pathologist. This finding suggests that our EBVNet could provide an innovative approach for the identification of EBVaGC and may help effectively select patients with gastric cancer for immunotherapy.

Deep learning model to predict Epstein–Barr virus associated gastric cancer in histology

“…Of the studies with weakly supervised learning, Muti et al reported a comparable performance with an AUROC of 0.859 on the external dataset. However, the results of these previous studies elucidate that supervised learning approaches, where the model was trained on tumor patches, outperformed the weakly supervised approach with an AUROC of 0.941 29 .…”

“…Until now, seven studies on EBV status prediction via a deep learning approach using digitalized WSIs have been published (Supplementary Table S1 ) 25 – 31 . Of these, the most similar to our pipeline are those proposed by Zhang et al and Zheng et al 28 , 29 ; in these two previous studies, a two-step approach utilizing a tumor classifier and an EBV classifier was implemented. Most studies have trained EBV classifiers using tumor patches generated from tumor annotation, with 28 , 29 or without 25 – 27 training tumor classifiers—a tumor annotation-based approach.…”

“…Of these, the most similar to our pipeline are those proposed by Zhang et al and Zheng et al 28 , 29 ; in these two previous studies, a two-step approach utilizing a tumor classifier and an EBV classifier was implemented. Most studies have trained EBV classifiers using tumor patches generated from tumor annotation, with 28 , 29 or without 25 – 27 training tumor classifiers—a tumor annotation-based approach. Other studies have trained the model without using tumor or normal patches—an annotation-free training approach 30 , 31 .…”

“…In the current study, we assumed that normal patches cannot provide any information about the genetic trait of the tumor. Therefore, we only used tumor patches based on the tumor annotation dataset when training the EBV classifier, similar to previous studies 25 – 29 . The tumor classifier in our proposed framework may have been useful in selecting representative patches (tumor patches) and in removing noise images of normal patches.…”

A comparative study of gastric histopathology sub-size image classification: From linear regression to visual transformer

“…In the field of medical image classification, deep learning algorithms are the most effective algorithms, and Convolutional Neural Network (CNN) is a widely used model for image classification, which can extract information from original medical images and classify normal and abnormal case images ( 17 ). Recently, Visual Transformer, which is originally applied to Natural Language Processing tasks, have become popular in computer vision, and Vision Transformer (ViT) have effective classification results when trained on large amounts of data and can significantly reduce the computer hardware and software resources required for training ( 18 ).…”