Interpretable MRI radiomics using optimal 1-mm peritumoral region predicts lymphovascular space invasion in cervical cancer

MRI radiomics model using intratumoral and 1-mm peritumoral features predicted lymphovascular space invasion in cervical cancer with AUC 0.843 (95% CI 0.731–0.955) in the test cohort; wider peritumoral expansions did not improve performance.

• Expanding the peritumoral region beyond 1 mm (3 or 5 mm) did not further enhance model discrimination.
• SHAP analysis indicated texture features dominated predictions, and adding clinical variables (tumor diameter, SCC level) offered no significant incremental value.

Automated summary

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