Vision transformer tops radiologist O-RADS 2022 performance for ovarian mass ultrasound, with hybrid AI-radiologist fusion yielding highest accuracy

For ultrasound-based ovarian mass risk stratification, a vision transformer (ViT16-384) achieved AUC 0.941 and accuracy 87.4%, significantly outperforming radiologists using O-RADS v2022 (AUC 0.683, accuracy 68.0%). Hybrid human-AI models further boosted performance.

• Radiologist-only O-RADS v2022 assessment achieved AUC 0.683 (accuracy 68.0%) on 512 ultrasound images from 227 patients.
• Best deep learning model (ViT16-384) reached AUC 0.941 and accuracy 87.4%, outperforming all CNNs.
• Hybrid models combining radiologist O-RADS scores with DL predictions significantly improved 9 of 12 CNNs and 3 of 4 ViTs (p<.05).

Automated summary

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