Artificial Intelligence Assisted Diagnosis for Peripheral Arterial Disease (AI-PAD)
Peripheral arterial disease (PAD) is a disease in which plaque builds up in the arteries, limiting the flow of oxygen-rich blood to the head, organs and limbs. A complete medical evaluation of PAD is based on computed tomography angiography (angio-CT). Interpreting an angio-CT is a complex medical examination, requiring good knowledge of anatomical variability among patients. It takes a lot of time, as the doctor must identify the route of each artery and analyze each image from a series of several thousands of images per patient. Currently, PAD is under-diagnosed and under-treated.
The challenge to be addressed is to automatically quantify the degree of disease in angio-CTs.
Given the complexity of the diagnostic process, the current project aimed at developing an intelligent software system able to automatically identify the arteries, detect abnormalities and facilitate the diagnosis. It is addressed to general radiologists, interventional radiologists, and vascular surgeons, aiming at reducing the diagnostic time, increasing the accuracy by highlighting the pathologies and providing accurate measurements (degree and length of stenosis), acting as a decision support system for preoperative planning.
The main results of the project consist in:
– The development of an AI module able to automatically, with no human intervention, segment bones, arteries, calcifications and thrombosis in angio-CTs; it is mainly based on computer vision and deep learning;
– The development of algorithms for reconstructing obstructed arteries, based on graph shortest path heuristics, to be applied as a post-processing step after segmentation;
– The implementation of a 3D reconstruction module that uses the extracted masks for bones, arteries, calcification, and thrombosis to generate STL images;
– The integration of the modules above within a software system for CT manipulation/inspection.