Multiparametric MRI radiomics differentiates tumor from treatment effects in glioblastoma

An MRI radiomics model using ADC, DSC, T1CE, and FLAIR distinguished tumor from treatment effects in post-treatment glioblastoma (AUC 0.887, sensitivity 75.0%, specificity 92.3%), but it requires external validation.

• SHAP analysis revealed balanced contributions from ADC, DSC, T1CE, and FLAIR sequences.
• Platt calibration improved model calibration (Brier score decreased from 0.229 to 0.132).
• The study was internally validated; external validation is necessary to confirm clinical utility.

Automated summary

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