MONDAY, Sept. 25, 2023 (HealthDay News) -- Unruptured cerebral aneurysms (UCAs) of sizes that warrant attention and possibly treatment that are missed in routine care can be identified by a machine learning algorithm, according to a study published online Sept. 13 in Stroke: Vascular and Interventional Neurology.
Hyun-Woo Kim, M.D., from UTHealth McGovern Medical School in Houston, and colleagues assessed the performance of a machine learning algorithm to identify UCAs among patients who underwent computed tomography angiography for evaluation of possible stroke. The images were analyzed using a convolutional deep neural network (Viz ANEURYSM) trained to identify UCAs ≥4 mm.
The researchers found that 4.2 percent of the 1,191 computed tomography angiograms performed during the study period were flagged by the machine learning algorithm as possibly demonstrating a UCA; 31 of these 50 cases were confirmed as true positive (positive predictive value, 62 percent). A total of 36 true aneurysms were identified, with four cases of multiple aneurysms. The internal carotid artery was the most common location (42 percent). Of the 36 cases, 10 and 24 were not noted in the clinical radiology report or clinical notes (median size, 4.4 mm) and were not referred for follow-up (median size, 4.4 mm), respectively. Fifteen of the 24 aneurysms not referred for follow-up had been noted in the radiology report. Overall, five of the 15 cases detected but not referred had a diameter greater than 7 mm.
"These findings suggest that a machine learning-based approach that can assist with screening for UCAs, and alerting the clinical care teams may improve the care of patients with UCAs," the authors write.
Several authors disclosed ties to medical device and technology companies, including Viz.ai., which funded the study.
Abstract/Full Text
