Authors: Chaima Dachraoui; Aymen Mouelhi; Salam Labidi
Addresses: Research Laboratory of Biophysics and Medical Technologies, Higher Institute of Medical Technologies of Tunis, University of Tunis El Manar, 1006, Tunis, Tunisia ' Laboratory of Signal Image and Energy Mastery, LR13ES03 (SIME), ENSIT, Tunis University, Tunis, Tunisia ' Research Laboratory of Biophysics and Medical Technologies, Higher Institute of Medical Technologies of Tunis, University of Tunis El Manar, 1006, Tunis, Tunisia
Abstract: Magnetic resonance imaging is considered a powerful tool for the no-invasive diagnosis of brain pathologies. Multiple sclerosis is an autoimmune inflammatory disease of the central nervous system in which clinical markers are used today for diagnosis and for therapeutic evaluation. In order to automate a long and hard process for the clinician, we propose a semi-automatic lesions segmentation approach in longitudinal image sequences. We use firstly a robust algorithm that allows spatiotemporal extraction by the geodesic active contour model. Then, we recommend an original scheme based on an automated image registration technique for evaluating the evolution of the detected lesions. A quantitative study is presented in this paper to validate our results using the 'BrainWeb' simulator, MICCAI2008, and MICCAI2016. Promising results are obtained in the case of clinical data. Our research was tested on ten typical and atypical synthetic motifs and 1,000 MR-images from different centres.
Keywords: multiple sclerosis lesions; brain MRI; segmentation approach; evolution; follow-up.
International Journal of Modelling, Identification and Control, 2021 Vol.38 No.1, pp.32 - 45
Received: 06 Jul 2020
Accepted: 18 Jan 2021
Published online: 04 Apr 2022 *