Improving AI model for interpretable brain MRI lesion detection

 4
Improving AI model for interpretable brain MRI lesion detection

The researchers included radiology report–derived textual features to evaluate AI model of Brain MRI lesion.

They evaluated 35 282 brain MRI scans and corresponding radiology reports using ReportGuidedNet model and compared it with PlainNet model without textual features,

The authors found that ReportGuidedNet outperformed PlainNet for all diagnoses indicating radiology report textual features improved the ability of the ReportGuidedNet model to detect brain lesions,

https://pubs.rsna.org/doi/10.1148/ryai.230520

https://sciencemission.com/Boosting-Deep-Learning-for-Interpretable-Brain-MRI-Lesion-Detection-through-the-Integration-of-Radiology-Report-Information