Biomarker identification for vertigo through human connectome analysis
Files
Anizet_35011700_Doyen_25731700_2022.pdf
Open access - Adobe PDF
- 23.44 MB
Details
- Supervisors
- Faculty
- Degree label
- Abstract
- Thanks to advances in magnetic resonance imaging (MRI), interest in the investigation of neurological disorders, such as some types of vestibular dysfunction, has increased. Vestibular system dysfunction can lead to vertigo, described as an abnormal sensation of motion. The objective of this work is to analyse diffusion and functional MRI images to identify connections between regions of the brain allowing to discriminate people suffering from vertigo from control participants. To achieve this goal, structural connectivity matrices, reflecting the fibers density between brain areas, and functional connectivity matrices, assessing the simultaneous activity between regions of the brain at rest, have been constructed. Statistical tests have been performed and two models commonly used in machine learning algorithms, the Support Vector Machine and the Random Forest, have been used. Anatomical connections between regions involved in cognitive, proprioceptive, and affective processing appear to be significant in discriminating control participants from vertigo participants. For both structural and functional features, regions located in the frontal lobe (such as the precentral gyrus and orbital gyrus), in the limbic lobe (like the cingulate gyrus) or the occipital lobe are likely to play a role in the vestibular system and its performances. To a lesser extent, regions in the temporal lobe (such as the fusiform, middle and superior temporal gyrus) and in the parietal lobe (such as the precuneus) also seem to contribute to vestibular processing. This should encourage researchers to study functional and structural connections of the vestibular system to strengthen the results of this work. Combined with a supplementary biological input from doctors, these findings could help for the diagnosis and treatment of vertigo patients.