Aortography keypoint tracking

Customer: Kemerovo Cardiology Center (Kemerovo, Russia)

Summary: Transcatheter aortic valve implantation (TAVI) is currently the most effective treatment option for patients with aortic stenosis. However, the success of TAVI procedures is highly dependent on the accuracy of valve positioning, which can be difficult to achieve with conventional imaging techniques. In response to these limitations, the development of new visual assistance systems is necessary. This study proposes an innovative multi-task learning algorithm to track the location of anatomical landmarks and identify critical key points on the aortic valve and delivery system during TAVI procedures. To improve the speed and accuracy of labeling, 9 neural networks were designed and tested to predict 11 keypoints of interest. These models were based on different neural network architectures, including MobileNet V2, ResNet V2, Inception V3, Inception ResNet V2, and EfficientNet B5. During training and validation, the ResNet V2 and MobileNet V2 architectures showed the best prediction accuracy/time ratio, with 97% and 96% accuracy and 4.7% and 5.6% mean absolute error, respectively. Based on this study, we believe that neural networks with these architectures can perform real-time predictions of aortic valve and delivery system position, thereby contributing to proper valve positioning during TAVI procedures.

Collaborators: Evgeny Ovcharenko (Research Institute for Complex Issues of Cardiovascular Diseases, Kemerovo, Russia), Kirill Klyshnikov (Research Institute for Complex Issues of Cardiovascular Diseases, Kemerovo, Russia)

Project type: Commercial / Research

Media: Journal paper, Interactive report, News 1 (rus), News 2 (rus), News 3 (rus)