AI-directed scan protocols cut back on unnecessary MRI sequences
Artificial intelligence could serve as an effective tool for determining which patients may need additional breast MRI sequences in real-time.
MRI offers the best visualization of breast tissue, especially in women with higher breast density. However, the modality can be time-consuming and costly, and it’s possible that not all patients require the full gamut of sequences in standard breast MRI protocols.
Research published this week in the journal Radiology suggests that AI could be the solution, as experts have determined it can be used as a triage tool to reduce the number of necessary sequences in real-time.
“MRI protocols typically involve many imaging sequences and often require too much time,” Sarah Eskreis-Winkler, MD, PhD, with the department of radiology at Memorial Sloan Kettering Cancer Center in New York, and colleagues noted. “Image-based artificial intelligence triage could be used to make modifications to breast MRI scanning protocols in real time, ensuring that sufficient diagnostic information is obtained for each patient as quickly as possible.”
For their research, the team applied an in-house AI tool to the breast MRI scans of nearly 900 women. The algorithm generated a suspicion score for subtraction maximum intensity projection images and determined whether patients needed to proceed with the full MRI protocol or if an abbreviated version (dynamic contrast-enhanced MRI scans only) would be sufficient instead. Patients with suspicion scores under the 50th percentile were read using both the AB-MRI protocol and the full MRI protocol.
Within 12 months of their exam, 51 of the 863 patients were diagnosed with cancer. The two protocols achieved similar performance in terms of sensitivity, specificity and accuracy. AI-directed abbreviated scans yielded a sensitivity of 88.2% versus 86.3% for the standard protocol, a specificity of 80.8% versus 81.4% and a positive predictive value of 23.6% compared to 24.7%.
In terms of detection rates, the AI-directed was projected to identify 31.6 per 1,000 exams, compared to 30.9 for the full protocol. Interval cancer detection was similar as well. The team determined that none of the AI-triaged scans would have resulted in cancer diagnoses had patients completed the full set of sequences instead.
“AI-directed stratified MRI decreased simulated scan times while maintaining diagnostic performance,” the team concluded.
In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She began covering the medical imaging industry for Innovate Healthcare in 2021.