Artificial Intelligence-Based Diagnostic System Differentiating Gastric Cancer and Ulcer: Comparison Between Original and Newly Developed System
Ken NAMIKAWA1, Toshiaki HIRASAWA1, Junko FUJISAKI1, Tomohiro TADA2, 3, 4
Affliations:
1Department of Gastroenterology, Cancer Institute Hospital, Japanese Foundation for Cancer Research, Tokyo, Japan
2AI Medical Service inc., Tokyo, Japan
3Tada Tomohiro Institute of Gastroenterology and Proctology, Saitama, Japan
4Department of Surgical Oncology, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan
Corresponding email address:
Ken NAMIKAWA (ken.namikawa@jfcr.or.jp)
Background and Aim: Previously, we were the first to report the usefulness of artificial intelligence (AI) systems for detecting gastric cancers (GC). However, the "original Convolutional Neural Network (O-CNN)" in the previous research had a relatively low positive predicted value (PPV). We aimed to develop a more accurate AI-based diagnostic system and evaluate its applicability for the differential diagnosis of GC and gastric ulcers (GU).
Methods: We constructed an "advanced CNN" (A-CNN) by adding new training-set data (4,453 GU images from 1172 lesions) into the O-CNN which was trained by 13,584 GC and 373 GU images. The diagnostic performances of the A-CNN and O-CNN were evaluated by validation dataset (739 images from 100 early GCs and 720 images from 120 GUs). All training and validation dataset were images without magnification.
Results: The sensitivity, specificity, and PPV of the A-CNN in classifying gastric cancer at the lesion level were 99.0% (95% confidence interval [CI] 94.6%−100%), 93.3% (95 %CI 87.3%−97.1%), and 92.5% (95%CI 85.8%−96.7%), respectively, and for classifying gastric ulcers were 93.3% (95%CI 87.3%−97.1%), 99.0% (95%CI 94.6%−100%), and 99.1% (95%CI 95.2%−100%), respectively. At the lesion level, the overall accuracies of the O- and A-CNN for classifying gastric cancers and gastric ulcers were 45.9% (gastric cancers 100%, gastric ulcers 0.8 %) and 95.9% (gastric cancers 99.0%, gastric ulcers 93.3%), respectively.
Conclusions: The AI-based diagnostic system could effectively not only to detect but also to perform the differential diagnosis of gastric cancers and ulcers.