- Synthesizing Contrast-Enhanced MR Images from Noncontrast MR Images Using Deep Learning
The authors developed and trained a novel deep learning model utilizing a diverse multi-institutional data set that was able to synthesize virtual contrast-enhanced T1-weighted images for primary brain tumors by using only noncontrast FLAIR, T2-weighted, and T1-weighted images.
- AI-Assisted Summarization of Radiologic Reports: Evaluating GPT3davinci, BARTcnn, LongT5booksum, LEDbooksum, LEDlegal, and LEDclinical
This study quantitatively evaluated the performance of natural language processing models using real, longitudinal brain aneurysm imaging reports to objectively understand the strength and weakness of these new technologies.
