Automated Detection of Necrotizing Soft Tissue Infection Features by Computed Tomography

Background/Objectives: To develop and evaluate an automated detection system for necrotizing soft tissue infection (NSTI) features on computed tomography (CT) images using the You Only Look Once version 10 (YOLOv10) model, aiming to improve diagnostic efficiency and surgical planning. Methods: This retrospective study included 31 patients with surgically confirmed NSTIs, spanning 2017-2023, from Chi Mei Medical Center, Taiwan. A total of 9001 CT images were annotated for four NSTI features: soft tissue ectopic gas, fluid accumulation, fascia edematous changes, and soft tissue non-enhancement. Model performance was evaluated using mean Average Precision (mAP), recall, and precision metrics. Results: The model achieved a mAP of 0.75, with recall and precision values of 0.74 and 0.72, respectively. Recall values for individual features were 0.76 for soft tissue ectopic gas, 0.66 for soft tissue non-enhancement, 0.92 for fascia edematous changes, and 0.68 for fluid accumulation. Conclusions: The YOLOv10-based system effectively detects four NSTI features on CT, including soft tissue ectopic gas, fluid accumulation, fascia edematous changes, and soft tissue non-enhancement.