Body / AbdominalAI / InformaticsResearch

MRI Radiomics Model Distinguishes Gleason ≤3+4 from ≥4+3 Prostate Cancer with External AUC 0.795

Radiology AI literature (PubMed)Jun 11

Combined clinical and peritumoral T2WI radiomics model: external validation AUC 0.795 (95% CI 0.694–0.895) for predicting Gleason ≥4+3 vs ≤3+4 in 225 patients (retrospective, multicenter). Core radiomic feature partially mediated the PI-RADS to Gleason score relationship (indire…

  • Multicenter retrospective cohort of 225 prostate cancer patients with T2-weighted MRI; intratumoral and peritumoral radiomic features extracted.
  • Best model integrated clinical variables and IntraPeri_1mm features; training AUC 0.975, internal test 0.854, external validation 0.795.
  • SHAP analysis revealed a core feature correlating with aggressive indicators and partially mediating PI-RADS effect on Gleason score.

Retrospective design; external AUC lower than training, consistent with possible overfitting; no comparison with radiologist interpretation.

Read the source

Related reporting systems

Automated summary

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