Automated CT-based sarcopenia assessment for risk stratification of patients undergoing colorectal cancer resection

Résumé fourni par la source

Despite the prognostic relevance of sarcopenia in colorectal cancer, it has not yet been incorporated into routine clinical patient assessment. This study investigates the potential of automatically CT-derived muscle-to-bone ratio (MBR) for preoperative stratification of colorectal cancer patients. We retrospectively analyzed CT images of 117 colorectal cancer patients undergoing surgical resection. A deep learning model was used to assess the abdominal MBR as a measure of sarcopenia. Univariable and multivariable analyses were performed to analyze the association between MBR and overall survival (OS), in-hospital mortality, length of stay (LOS), and postoperative C-reactive protein (CRP) levels. In univariable analysis, preoperative MBR was significantly associated with OS (hazard ratio (HR) 0.29, 95 ​% CI: 0.13–0.64, p ​< ​0.005). In multivariable analysis adjusted for age, sex, and UICC stage, higher MBR remained independently associated with improved OS (HR 0.28, 95 ​% CI: 0.10–0.79, p ​= ​0.017) and reduced in-hospital mortality (coefficient (β) ​= ​-2.58, p ​= ​0.031). Subgroups based on MBR showed significantly different OS in Kaplan-Meier analysis (p ​< ​0.005). Furthermore, patients with low preoperative MBR exhibited significantly higher postoperative CRP values (p ​= ​0.039). No significant association was observed between MBR and LOS. Our study demonstrates the potential of deep learning-derived MBR for automated sarcopenia assessment and patient stratification in colorectal cancer surgery.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.