Thursday, December 4 | 8:40 a.m.-8:50 a.m. | R1-SSCA09-5 | Room E353C
An AI algorithm shows potential for utilizing routine chest CT exams to opportunistically predict patient risk of having a major adverse cardiac event (MACE), according to this scientific presentation.
Traditional methods for estimating the risk of MACE utilize only known clinical and demographic risk factors; radiology images aren’t incorporated directly. However, these risk estimation techniques perform suboptimally when applied to external institutions with shifting clinical and demographic distributions, according to presenter Amara Tariq, PhD, of the Mayo Clinic College of Medicine and Science.
In an effort to improve this situation, the group developed a 3D convolutional neural network (CNN) to estimate MACE risk by analyzing chest CT exams. An approach called “causal intervention” was utilized to enhance the model’s generalizability to external patient populations.
The best-performing CNN achieved an area under the curve of 0.73 on the internal test set and 0.69 on the external test set, outperforming current models and generalizing well to external data with significant shift in patient populations, according to the researchers.
“Leveraging AI to analyze rich, underutilized data within routinely performed chest CTs holds significant promise for enhancing early MACE risk estimation across diverse patient populations beyond the capabilities of current clinical tools,” the authors wrote.
Learn more about the power of causal intervention for improving AI model generalizability in this Thursday morning presentation.
