Prediction of invasion depth and lymph node metastasis in early gastric cancer by artificial intelligence and Gene expression profiling

2020 Fiscal Year Final Research Report

• Project/Area Number: 19K16825
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 50020:Tumor diagnostics and therapeutics-related
• Research Institution: The University of Tokyo
• Principal Investigator: Kataoka Yosuke 東京大学, 医学部附属病院, 届出研究員 (80800896)
• Project Period (FY): 2019-04-01 – 2021-03-31
• Keywords: 早期胃癌 / 深達度 / リンパ節転移リスク / AI / 網羅的遺伝子解析

Outline of Final Research Achievements

As for the diagnosis of gastric cancer invasion depth by AI, 1207 endoscopic images of intramucosal cancer, 1048 images of submucosal slight invasive cancer (<500 μm), 1260 images of submucosal massive invasive cancer (>500 μm), and 716 images of advanced cancer were annotated. 75% of the dataset was used for training and 25% for evaluation. AI showed that the positive detection rate was 67% with the AUC value of 0.76.
For the gene expression profiling, RNA from 8 cases of intramucosal cancer and 8 cases of submucosal massive invasive cancer were extracted for microarray analysis. We focused on gene A with large variation among the two groups. Through the immunohistological examination, the mean number of positive cells in a 400-fold field of view for gene A was 3.8 for intramucosal cancer and 0.4 for submucosal massive invasive cancer with a significant difference (p < 0.001). The ROC curve for the diagnosis of invasion depth using gene A showed that the AUC value was 0.71.

Free Research Field

消化器内科

Academic Significance and Societal Importance of the Research Achievements

早期胃癌の深達度及びリンパ節転移リスクに関する術前診断は内視鏡的粘膜下層剥離術の適応判定および追加外科切除を減らす上で重要である。これまで胃癌深達度診断は、内視鏡医が肉眼所見をもとに主観的評価よって行われてきたが、本研究ではAIおよび網羅的遺伝子発現解析による新たな予測法の可能性を示した。