CT radiomics and deep learning of intra- and peri-thrombus regions predict functional outcome after thrombectomy

A late-fusion model combining CT radiomics and deep learning features from intra- and peri-thrombus regions predicted 90-day functional outcome after mechanical thrombectomy with an AUC of 0.836 (95% CI 0.769–0.891) in an external test cohort.

• Retrospective multicenter study: 315 patients training (Center A) and 156 external test (Centers B and C).
• Late-fusion combined model significantly outperformed peri-thrombus alone (AUC 0.758, p=0.005) and early-fusion (AUC 0.763, p=0.009).
• Model improved reclassification: NRI 0.935 and IDI 0.278 compared with intra-thrombus model.

Automated summary

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