Abstract:

Acute aortic pathologies, particularly dissections and aneurysms, are life-threatening 
conditions that require rapid and precise diagnosis. Traditional manual analysis of computed 
tomography (CT) images is often associated with subjectivity and variability. The aim of the present 
study is to develop a hybrid artificial intelligence system for automated morphometric analysis that 
can serve as a reliable digital assistant in clinical decision-making. The methodology integrates 
classical computer vision algorithms with modern deep convolutional neural networks (U-Net 
architecture) within a data-centric philosophy focused on data quality. The results obtained from 
the developed prototype demonstrate successful semantic segmentation of aortic lumens and intimal 
flaps, as well as the elimination of common CT artifacts. The conclusions emphasize that, at an 
early stage of development, priority should be given to the quality of annotated data, and that the 
system should complement rather than replace the physician, contributing to effective human
artificial intelligence collaboration in medical imaging diagnostics. 

Keywords: aortic pathologies, computed tomography, artificial intelligence, semantic segmentation, 
medical image analysis. 

Download the full article here: Download 2017