Artificial intelligence

Our research group is engaged in a spectrum of artificial intelligence research, including the reconstruction of medical images, image correction, and the segmentation of anatomical structures from cardiac images.

Our research spans medical imaging modalities, with a focus on computed tomography (CT), 4D Flow magnetic resonance imaging (MRI) and Dixon MRI. While these imaging modalities hold great promise, they often demand substantial computational resources. Fully automatic methods would eliminate the risk of observer bias in the complicated analysis pipeline, as well as greatly simplify and accelerate the time-consuming image acquisition, reconstruction, and image processing. As an example, a classic acquisition and reconstruction of a complete 4D flow MRI dataset on the scanner could take up to an hour, while AI based methods can achieve it in the order of seconds. Similarly, automatic segmentation of the heart from CT data can be achieved in under a minute while manual identification of the cardiac structures can take hours in a CT dataset which is typically comprised of hundreds of individual slices.

Additionally, our group is investigating novel clinical tools made possible through artificial intelligence. This work includes, for example, head-to-head comparisons of CT and MRI and image-to-image translation for image denoising and artifact correction using generative AI.