Multisequence MRI fusion model predicts 24-month glioma survival with AUC 0.899

Retrospective study: end-to-end multisequence MRI fusion model (EMFA) achieved AUC 0.899 and 93.5% accuracy for preoperative 24-month glioma survival prediction, outperforming single-sequence baselines.

• Retrospective EMFA model integrates deep transfer learning and radiomics features for glioma survival prediction.
• Optimal input: nine channels for T1w and T2w, three channels for contrast-enhanced T1w MRI.
• Model outperformed single-sequence baselines; AUC 0.899, accuracy 93.5% in held-out test cohort.

Retrospective design; sample size and external validation not detailed in abstract.

Automated summary

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