Hybrid MRI radiomics and deep learning model preoperatively identifies E. coli in perianal abscess

A hybrid MRI radiomics-deep learning nomogram predicted E. coli in perianal abscesses with AUC 0.885, sensitivity 81.8%, specificity 81.2%, potentially guiding preoperative antibiotics.

• Retrospective study of 215 patients with culture-confirmed perianal abscesses used multi-sequence MRI (T1WI, T2WI, FS-T2WI).
• Gender (OR 0.456, p=0.001) and diabetes (OR 9.207, p<0.001) were independent predictors of E. coli.
• Limitation: retrospective, single-center design without external validation.

RadPigeon summaries are original and for information only. They are not clinical advice.