Monday, December 2 | 9:30 a.m.-9:40 a.m. | M3-SSNR03-1 | Room S402
Here, attendees learn about the performance of a deep learning-based ultrasound method that can predict stroke risk.
In his talk, Saman Doroodgar Jorshery, MD, from Harvard Medical School in Cambridge, MA will present findings showing how imaging findings are tied to incident stroke and reduced brain gray matter.
The researchers developed a convolutional neural network (CNN) model called Carotid-Risk, which used carotid ultrasound images of 59,137 participants in the U.K. Biobank, a population study of volunteers in the U.K.
They trained the model on 80% of the dataset to predict biological age. This refers to a person’s chronologic age at the ultrasound time-adjusted for cardiovascular risk factors, comorbidities, all-cause mortality, and major cardiovascular events during follow-up.
For the test set (n = 11,832), the team evaluated potential ties between risk scores given by Carotid-Risk with observed stroke incidents over a median of 3.8 years and brain volume of cortical gray matter. The latter was derived from T1-weighted brain MR images.
In the test set, after adjusting for covariates, the team found that each standard deviation increase in Carotid-Risk was significantly linked to incident stroke. This included an adjusted hazard ratio (aHR) of 1.37 (p = 0.03).
In the subset of testing dataset participants with brain MRI (n = 7,559), after adjusting for all covariates, the researchers observed that each standard deviation increase in Carotid-Risk was significantly associated with a 0.05 standard deviation decrease in cortical gray-matter volume. Finally, there was no significant association between Carotid-Risk and white-matter volume (p = 0.27).
The researchers highlighted that this model could help predict stroke beyond chronological age and cardiovascular risk factors.
“Deep learning derived Carotid-Risk may help direct preventive interventions to those at high-risk for neurovascular disease and improve understanding of the role of the carotid artery in neuronal degeneration,” they wrote.
Find out more by attending this session.
