Establishment of a Prognostic Prediction System for Hepatocellular Carcinoma after Resection Using EOB-MRI and Deep Learning

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K07647
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 52040:Radiological sciences-related
• Research Institution: Kumamoto University
• Principal Investigator: Nakagawa Masataka 熊本大学, 大学院生命科学研究部(医), 特定研究員 (30771125)
• Co-Investigator(Kenkyū-buntansha): 山下 洋市 熊本大学, 大学院生命科学研究部(医), 准教授 (00404070) ; 川上 史 熊本大学, 病院, 特任助教 (40565678) ; 三上 芳喜 熊本大学, 病院, 教授 (90248245) ; 中浦 猛 熊本大学, 大学院生命科学研究部(医), 准教授 (90437913)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: 放射線診断学 / 肝細胞癌 / Deep Learning / MRI / 予後予測

Outline of Final Research Achievements

In this study, we attempted to establish a prognostic prediction system for hepatocellular carcinoma (HCC) after resection using EOB-MRI images and Deep Learning. We investigated whether Deep Learning could predict prognostic factors such as differentiation grade (low differentiation: 47 cases, high/moderate differentiation: 133 cases) using EOB-MRI images. However, we found that the system was highly dependent on the scanner model. After examining various preprocessing and Deep Learning methods, we found that a certain level of prediction was possible when using the same scanner model, but the performance significantly decreased when tested on a different scanner model. Specifically, the prediction performance was inferior to conventional methods such as ADC values. Furthermore, we also investigated liver function, which greatly influences patient prognosis, but it was difficult to achieve novelty compared to conventional methods.

Free Research Field

放射線科学関連

Academic Significance and Societal Importance of the Research Achievements

今回の検討では、当初の目論見であったEOB-MRIとDeep Learningを用いたHCC切除後の予後予測システムを確立することができませんでした。これは、EOB-MRI 画像を用いた予後予測における機種依存性の高さが原因と考えられ、これを克服するためにはさらなる技術革新が必要である事がわかりました。