Histopathological image analysis is an essential process for the discovery of diseases such as cancer. However, it is challenging to train CNN on whole slide images (WSIs) of gigapixel resolution considering the available memory capacity. Most of the previous works divide high resolution WSIs into small image patches and separately input them into the model to classify it as a tumor or a normal tissue. However, patch-based classification uses only patch-scale local information but ignores the relationship between neighboring patches. If we consider the relationship of neighboring patches and global features, we can improve the classification performance. In this paper, we propose a new model structure combining the patch-based classification model and whole slide-scale segmentation model in order to improve the prediction performance of automatic pathological diagnosis. We extract patch features from the classification model and input them into the segmentation model to obtain a whole slide tumor probability heatmap. The classification model considers patch-scale local features, and the segmentation model can take global information into account. We also propose a new optimization method that retains gradient information and trains the model partially for end-to-end learning with limited GPU memory capacity. We apply our method to the tumor/normal prediction on WSIs and the classification performance is improved compared with the conventional patch-based method.
To overcome the increasing burden on pathologists in diagnosing gastric biopsies, we developed an artificial intelligence‐based system for the pathological diagnosis of gastric biopsies (AI‐G), which is expected to work well in daily clinical practice in multiple institutes. The multistage semantic segmentation for pathology (MSP) method utilizes the distribution of feature values extracted from patches of whole‐slide images (WSI) like pathologists’ “low‐power view” information of microscopy. The training dataset included WSIs of 4511 gastric biopsy tissues from 984 patients. In tissue‐level validation, MSP AI‐G showed better accuracy (91.0%) than that of conventional patch‐based AI‐G (PB AI‐G) (89.8%). Importantly, MSP AI‐G unanimously achieved higher accuracy rates (0.946 ± 0.023) than PB AI‐G (0.861 ± 0.078) in tissue‐level analysis, when applied to the cohorts of 10 different institutes (3450 samples of 1772 patients in all institutes, 198–555 samples of 143–206 patients in each institute). MSP AI‐G had high diagnostic accuracy and robustness in multi‐institutions. When pathologists selectively review specimens in which pathologist’s diagnosis and AI prediction are discordant, the requirement of a secondary review process is significantly less compared with reviewing all specimens by another pathologist.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.