November 17, 2026
Liu B, Fereydooni A, Jensen M, Lakshmi A, et al. AI-Driven Multimodal Risk Assessment Combining CT Imaging Biomarkers and Frailty Scores for Enhanced Mortality Prediction in Surgery Patients. J Am Coll Surg. October 2026.
Liu et al. examined whether AI-derived measures from existing CT scans could improve prediction of 1-year mortality after surgery when added to a diagnosis code-based frailty score. The single-center retrospective study included 7,672 patients who underwent nonemergency surgery within 6 months of an abdominal and pelvic CT scan. Overall, 12.3% died within 1 year of surgery, and 7,638 patients were included in the primary analysis.
The researchers combined the Risk Analysis Index, calculated from diagnosis codes (RAI-ICD), with automated CT measures of muscle, fat, bone, and aortic calcification. The frailty score performed better than any individual imaging measure. Adding the imaging measures together increased the area under the receiver operating characteristic curve from 0.75 to 0.79 (P < .001). The combined model’s predictions also aligned more closely with observed mortality, particularly among patients with higher predicted risk.
The findings suggest that CT scans already obtained during care could provide additional information for preoperative risk discussions. The authors called for external validation before clinical use.