Detecting atherosclerotic plaque calcifications of the carotid artery through optimum-path forest

• Main Author: Jodas, Danilo Samuel [UNESP]
• Publication Date: 2022
• Other Authors: Roder, Mateus [UNESP], Pires, Rafael [UNESP], Silva Santana, Marcos Cleison [UNESP], de Souza, Luis A., Passos, Leandro Aparecido [UNESP]
• Format: Book part
• Language: eng
• Source: Repositório Institucional da UNESP
• Download full: http://dx.doi.org/10.1016/B978-0-12-822688-9.00014-1; http://hdl.handle.net/11449/241373
• topic: Atherosclerotic lesions; Carotid artery; Computer-aided diagnosis; Medical images; Optimum-path forest

Analysis of the atherosclerotic lesions deposited in the carotid artery is so far an essential task for estimating possible cardiovascular disorders in patients. The immediate assessment of such lesion, as well as its morphology and composition, turns out necessary to avoid its progression beforehand, thus preventing more severe conditions such as heart attacks and strokes caused by calcified elements observed in advanced stages. Heretofore, a number of works addressed medical diagnosis problems through computational approaches, developing Computer-Aided Diagnosis (CAD) tools to detect, among several applications, atherosclerotic plaques formed in carotid arteries. In this context, a graph-based machine learning framework called Optimum-Path Forest (OPF) was successfully employed to tackle several CAD-based problems, even though no one still explores the model to classify the task mentioned above. Therefore this paper proposes the classification of regions in atherosclerotic lesions as calcified or noncalcified debris through OPF-based approaches. In the process, handcrafted features are extracted from pixels of computed tomography angiography images of the carotid artery. Also, each pixel is labeled by an expert as a calcified or noncalcified element. Thereafter, the OPF classifier, as well as four variants, namely Fuzzy OPF, OPF. knn, Probabilistic OPF, and the OPF for anomaly detection, are compared for the task of predicting whether the pixel of the carotid artery stands for the calcium of the atherosclerotic lesion or not. © 2022 Copyright