AI detects missed aneurysms in MR angiography with increased sensitivity
A team of Japanese researchers found that a deep learning-based algorithm used to analyze time-of-flight (TOF) MR angiography images improved cerebral aneurysm detection by an average of 8.9 percent and achieved an average sensitivity of 92 percent compared to initial radiology reports, according to research published Oct. 23 in Radiology.
TOF MR angiography can detect cerebral aneurysms with roughly 90 percent accuracy, however, the challenge is that aneurysms can differ in appearance depending on the type of imaging modality used and the conditions under which the TOF MR angiography was performed, wrote Daiju Ueda, MD, of the Osaka City University Graduate School of Medicine in Japan, and colleagues. In this case, deep learning may have the potential to reduce the number of overlooked aneurysms and could add value to a radiologist’s initial assessment.
Ueda and colleagues collected TOF MR angiography images from four medical institutions to develop the algorithm. The images were divided into three data sets—training data set, internal test data set and external test data set. Researchers evaluated the algorithm's sensitivity in detecting aneurysms from the images.
Two radiologists also performed a blinded interpretation of aneurysm individuals detected by the algorithm to find aneurysms that had been overlooked in the initial reports, according to the researchers.
A total of 748 aneurysms from 683 TOF MR angiography exams were identified—with 318 examinations performed on male patients (average age 63 years) and 365 on female patients (average age 64 years).
“Test data were provided by the internal test data set (649 aneurysms [mean size, 4.1 mm ± 3.2] in 521 examinations, including 177 male patients and 344 female patients with mean age of 66 years ± 12 and 67 years ± 13, respectively) and the external test data set (80 aneurysms [mean size, 4.1 mm ± 2.1] in 67 examinations, including 19 male patients and 48 female patients with mean age of 63 years ± 12 and 68 years ± 12, respectively),” the researchers wrote.
Overall, the algorithms demonstrated a sensitivity of 91 percent and 93 percent for the internal and external test data sets. It also showed an improved aneurysm detection rate in the both data sets by 4.8 percent and 13 percent, respectively compared to the initial reports.
The researchers noted the results also demonstrated highest detection rates for new aneurysms in the vertebral artery area and the basilar artery area for the location of cerebral aneurysms detected in the internal test data set.
“Radiologists sometimes fail to pay attention to the vertebral artery area and the basilar artery area because they are not common areas for aneurysms, so, with the support of the algorithm, the detection rates in these areas (given the epidemiology of aneurysm location) were high,” Ueda et al. wrote. “Our algorithm may improve the interpretation of cerebral aneurysms, helping to support everyday diagnostic tasks.”
A recent graduate from Dominican University (IL) with a bachelor’s in journalism, Melissa joined TriMed’s Chicago team in 2017 covering all aspects of health imaging. She’s a fan of singing and playing guitar, elephants, a good cup of tea, and her golden retriever Cooper.