Development of stopping criterion and cluster number selection strategies for an iterative student's t mixture model used for segmentation of multiple sclerosis lesions in magnetic resonance images

• Grant number: 19/25425-0
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): February 01, 2020
• Effective date (End): January 31, 2021
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Ricardo José Ferrari
• Grantee: João Gabriel Coli de Souza Monteneri Nacinben
• Host Institution: Centro de Ciências Exatas e de Tecnologia (CCET). Universidade Federal de São Carlos (UFSCAR). São Carlos , SP, Brazil

Abstract

Multiple Sclerosis (MS) is acronical inflammatory disease, possibly autoimmune, that impairs the Central Nervous System (CNS) and mainly affects the young adult population. Due to genetic or environmental factors, in a MS case, the immune system attacks neuron's myelin sheath, thus compromising CNS' function. Magnetic Resonance Imaging (MRI) has been clinically used with great success for the diagnosis and monitoring of MS, mainly due to the excellent contrast between soft tissues. In recent years, a few computational methods have been proposed to assist in the segmentation and volumetric measurement of MS lesions, among which we can mention a method recently developed by our group that is iterative, non-supervised and is based on Student's-t distributions mixture model. Notwithstanding the advantages of being automatic, unsupervised and presenting results comparable to other methods proposed in the literature, our method lacks a stopping criterion for the model, which is iterative but uses a fixed number of iterations, and also a criterion for selecting the number of clusters, which is, likewise, fixed. As possible solutions to the aforementioned limitations, in this research we will investigate using texture patterns of lesion maps, obtained at each iteration of the method, to define a stopping criterion for the algorithm. In addition, we will also analyze the Bayesian Information Criterion (BIC) and Akaike Information Criterion (AIC) as a way to determine the number of clusters for the mixture model.