Advances in AI: July 2023

A fully convolutional data description (FCDD) model for identifying anomalies on breast MRI demonstrated an 84 percent AUC for detection tasks in a balanced cohort with a 20 percent malignancy prevalence and a 72 percent AUC for detection tasks in an imbalanced group with a 1.85 percent cancer prevalence.

In a second part of a new podcast episode on recently published research on projected radiation-induced cancers from computed tomography (CT) scans, Mahadevappa Mahesh, MS, Ph.D., and Joseph Cavallo, M.D., offer current perspectives on cardiac CT dosing, AI advances and the importance of teamwork in ensuring appropriate dosing for CT.

The updated capabilities of SwiftMR include personalized scan settings within the software, artifact reduction and cloud integration.

In a new podcast, Mahadevappa Mahesh, MS, Ph.D., and Joseph Cavallo, M.D., share their perspectives on recently published research looking at projections for future radiation-induced cancers from computed tomography (CT) scans.

For patients with unresectable hepatocellular carcinoma, a pre-treatment MRI finding of LI-RADS tumor in vein was associated with over an 86 percent lower likelihood of responding to transcatheter arterial chemoembolization (TACE) and targeted immunotherapy.

For patients who had coronary CTA for acute chest pain, emerging research found no significant association between plaque burden grades with CAD-RADS 2.0 and cardiac events.