Chronic total occlusions (CTOs) are completely blocked blood vessels that contribute to more than 100,000 deaths and 6 million non-fatal complications in Europe every year. There is currently no system that can accurately guide a guidewire through complex CTO lesions. In APOLLO, we focus on a sub-problem of this by training AI models based on a new 3D dataset to be developed for the control of robotic systems.
The concrete innovation of APOLLO is the development and operationalization of AI-based technology to enable real-time navigation in complex blood vessels. We compile an annotated dataset consisting of point clouds (3D, volumetric data) of human vascular systems. We then train and validate AI models based on the annotated dataset. The primary focus of these models is segmentation: the specific classification of the different volumes within a point cloud. This information can then be used to determine the volume of a blockage and calculate the path to it through a blood vessel. We make the results available in accordance with open science and open data (FAIR) principles.