3Dボディ構成と人工知能—腹部緊急手術におけるサルコペニアの評価と術後転帰予測のための新規ツール
Chui Foong Ong1, Ke Cao1, Gabriel Lirios1
1Department of Surgery, Western Precinct, University of Melbourne, Victoria, Australia.
Backgrounds:
Sarcopenia is associated with higher mortality and morbidity in emergency laparotomies. Sarcopenia is traditionally measured with single 2D axial computed tomography (CT) slice at the L3 level, which is time-consuming and provide limited data. This study aims to determine if sarcopenia, measured using Artificial intelligence (AI) 3D-derived body composition (BC), can predict adverse outcomes after emergency abdominal surgery.
Methods:
Retrospective analysis of Australian and New Zealand Emergency Laparotomy Audit-Quality Improvement (ANZELA-QI) patients treated at a tertiary Australian hospital from 2018 to 2023 was conducted. Multiple CT slices from lumbosacral regions were used for 3D BC analysis using a validated AI segmentation model. Sarcopenia was defined based on the lowest quartile for skeletal muscle radiodensity.
Results:
408 patients were included. Sarcopenic patients had lower skeletal muscle mass (< 0.001) with higher volumes of visceral adipose tissue (p < 0.001) and subcutaneous adipose tissue (p < 0.02). Sarcopenia was associated with age (73 vs. 57 years; p < 0.001), increased length of stay (26 vs. 15 days; p = 0.041) and intensive care unit admission (p < 0.001). Sarcopenia was not associated with significant post-operative complications (Clavien-Dindo ≥ 3) (p = 0.903) or worse discharge status (p = 0.138).
Conclusion:
Sarcopenia is a significant predictor of adverse postoperative outcomes in patients undergoing emergency abdominal surgery. CT-derived 3D lumbosacral BC may help identify high-risk patients to guide risk stratification. AI has the potential to aid future implementation of 3D BC into routine clinical application.
