In order to develop an AI model that can accuratelycompletely segment coronary arteries, our team, the TW-CVAI, has established a training dataset composed of strictly verified annotations of coronary lumen boundaries in coronary CT angiography (CCTA). We designed a deep learning model, two-channel 3D-UNet, with a priori prerequisite (vesselness prior) to facilitate identification of vascular structures. The final model, the TaiCAD-Net, greatly shortens the CCTA interpretation time from 6 hours to 10 minutes, with the overall segmentation accuracy of 86 by Dice similarity coefficient.

The model, TaiCAD-Net, combines both deep learningmachine learning approaches to achieve "exploratory" image segmentation. This algorithm is able to recognize more coronary artery branches with the aid of vesselness prior,to delineate boundary precisely by patch-wise learning. The model currently achieves segmentation accuracy of 86 by Dice similarity coefficient. An additional 159 mm coronary arteries (8.2 branches) are identified. We won the third prize in the MICCAI 2020 automatic coronary artery segmentation accuracy global competition (ASOCA).

Coronary CT angiography (CCTA) has been recommended as the first-line diagnostic modality for coronary artery disease. Our TaiCAD-Net greatly shortens the CCTA interpretation time from 6 hours to 10 minutes, with high segmentation accuracy. The TaiCAD-Net can benefit manufacturers of computer tomography scanners (as an analytic module), hospitalshealth examination businesses (to provide reliable image interpretation), insurance companies,patients with cardiovascular diseases. This innovation makes the diagnosis of coronary artery disease readily available.

天使投資人、策略合作夥伴