A combined MRI/ultrasound model improves breast disease detection

Sunday, December 1 | 11:45 a.m.-12:15 p.m. | S3A-SPBR-7 | Learning Center

Compared with ultrasound or MRI used alone to diagnose breast diseases, a combined model of both modalities based on radiomics more accurately identifies breast conditions, according to a poster to be displayed Sunday morning.

Presenter Dong Bai, PhD, of Shanghai University of Traditional Chinese Medicine in Shanghai, China, and colleague Zhiqun Wang, PhD, also of the university, found that this multimodality imaging approach makes up for the "shortcomings of single examination methods, and [when combined with] radiomics [helps] avoid the mistakes of manual judgment."

Bai and Wang conducted a study that included 131 patients with breast conditions confirmed by pathology between January 2021 and August 2023. Of these patients 73 had benign breast conditions and 58 had malignant ones. They extracted radiomics features from the MRI and ultrasound exams to create an ultrasound model (20 features), an MRI model (11 features), and a combined model (14 features); they then created training and validation groups for each model and evaluated them for the area under the receiver operating characteristic (AUROC), sensitivity, specificity, and accuracy.

Overall, the researchers found the ultrasound/MRI combination model had the best diagnostic performance. It showed the following for the training and validation groups:

• An AUROC of 0.92 and 0.91
• A sensitivity rate of 80% and 67%
• A specificity rate of 90% and 94%
• An accuracy rate of 84% and 79%

"[The multimodal model provides] a better basis for clinical diagnosis and treatment," Bai and Wang wrote.

Head over to Lakeside Center on Sunday to learn more.