Predicting the Local Response of Esophageal Squamous Cell Carcinoma to Neoadjuvant Chemoradiotherapy by Radiomics with a Machine Learning Method Using 18F-FDG PET Images

2021 Fiscal Year Final Research Report

• Project/Area Number: 19K08153
• 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: Hiroshima University
• Principal Investigator: Murakami Yuji 広島大学, 医系科学研究科(医), 准教授 (10403528)
• Co-Investigator(Kenkyū-buntansha): 河原 大輔 広島大学, 医系科学研究科(医), 助教 (20630461) ; 西淵 いくの 広島大学, 病院(医), 講師 (70595834) ; 三木 健太朗 杏林大学, 保健学部, 准教授 (90732818)
• Project Period (FY): 2019-04-01 – 2022-03-31
• Keywords: 食道扁平上皮癌 / 化学放射線療法 / 治療効果予測 / Radiomics / 機械学習 / ニューラルネットワーク

Outline of Final Research Achievements

Using Radiomics and machine learning, we developed a model to predict pathological findings after preoperative chemoradiotherapy in patients with resectable locally advanced esophageal squamous cell carcinoma. A neural network was used as the machine learning method, and a five-fold cross validation analysis was performed to improve the accuracy. The mean predictive accuracy, specificity, sensitivity, and AUC of the model were 91.2%, 92.0%, 89.5%, and 0.97, respectively, which are very good results. The contents of this study were presented at the 33rd Annual Meeting of the Japanese Society for High Precision Torso Irradiation and the 62nd Annual Meeting of the American Society for Radiation Oncology. The study was also completed and accepted for publication in an international journal.

Free Research Field

放射線腫瘍学

Academic Significance and Societal Importance of the Research Achievements

切除可能局所進行食道扁平上皮癌に対する術前化学放射線療法による病理学的完全奏効率は比較的高く、治療前の情報から病理所見を予測できれば、臓器温存希望患者の治療選択に貢献できる。本研究では術前画像情報（PET画像）から高精度に術前治療（化学放射線療法）の病理学的完全腫瘍消失を予測するモデルが構築できた。さらに精度を高めることで、病理学的腫瘍完全消失が高率に予測される食道扁平上皮癌患者における臓器温存治療の選択根拠を与えうるツールとなりえる。今後、多施設によるさらに多くの症例数を用いたより汎用的なモデルの開発が望まれるが、本研究はその礎となると考える。