Multimodal MRI deep learning model shows high NPV for meningioma bone invasion in external validation

A multimodal MRI deep learning model achieved an AUC of 0.82 (95% CI 0.73–0.91) and NPV of 96% for preoperative prediction of meningioma bone invasion in external testing, improving over conventional radiomics (AUC +0.155, p=0.035).

• The model improved AUC by 0.155 (p=0.035) over conventional radiomics.
• SHAP analysis showed low-ADC features contributed dominantly to predictions.
• Decision-curve analysis confirmed net clinical benefit at a 20% risk threshold, driven by high NPV.

Automated summary

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