Gene mutation/expression prediction from clinical images using artificial intelligence technology in pancreatic cancer

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K17570
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 55010:General surgery and pediatric surgery-related
• Research Institution: Chiba Cancer Center (Research Institute)
• Principal Investigator: Iwatate Yosuke 千葉県がんセンター(研究所), 肝胆膵外科, 医長 (10815731)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: 膵臓癌 / 膵管腺癌 / Radiogenomics / ITGAV / ITGB1

Outline of Final Research Achievements

As a radiogenomics analysis, RNA-seq was performed on 12 pancreatic cancer tissues for an exhaustive search of genes useful for clinical pathology. A gene expression prediction model was constructed by machine learning using immunostaining (IHC) and CT images of 107 cases. ITGB1 and ITGAV were identified and significantly correlated with IHC (r=0.552 P=0.118, r=0.625 P=0.039). The prognosis worsened in the IHC high expression group (P=0.035, 0.009). Recurrence was similar (P=0.028, 0.003). The predictive model for ITGAV showed a certain detectability (AUC=0.697), and the group predicted to have high expression also had a poor prognosis (P=0.048).

Free Research Field

肝胆膵外科

Academic Significance and Societal Importance of the Research Achievements

診断時に進行していることが多く、5年生存率が10％程度と予後不良な膵臓癌において、診断のための侵襲的な生検等の検査や外科治療は専門施設やハイボリュームセンターでないと施行できない場合が多い。今回CTによる予後予測や今後治療の一助のなりうる可能性がある遺伝子マーカーの同定およびその発現予測が可能であった。今後症例を重ねることでCT画像等の臨床画像から非侵襲的に予後や治療マーカーの同定が行うことができれば、高額で時間がかかる遺伝子検査でなく簡易で費用対効果が高く、専門機関でなくても比較的短期間で結果が得られる検査となることも考えられる。