Segmentation of neonatal magnetic resonance imaging: a structural approach

• Grant number: 18/07386-5
• Support type: Scholarships in Brazil - Master
• Effective date (Start): July 01, 2018
• Effective date (End): July 29, 2020
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Cooperation agreement: Coordination of Improvement of Higher Education Personnel (CAPES)
• Principal researcher: Roberto Marcondes Cesar Junior
• Grantee: Larissa de Oliveira Penteado
• Home Institution: Instituto de Matemática e Estatística (IME). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Associated research grant: 15/22308-2 - Intermediate representations in Computational Science for knowledge discovery, AP.TEM
• Associated scholarship(s): 19/16112-9 - Graph learning for MRI semantic segmentation, BE.EP.MS

Abstract

Increasingly, it has been highlighted the importance of imaging exams for the diagnostic and follow-up of pathologies and the assessment of anatomical structures development. Thereby, the analysis of Magnetic Resonance Imaging of neonates and preterms brains is essential for the study of the patients' neurodevelopment, as well as the prognostic and prediction of possible future diseases. The manual analysis of MRIs is time-consuming, requires an expert with years of experience and is subject to variability according to the radiologist (that is, it is subjective). An alternative to it is the automatic segmentation of such images, which present some specific challenges: the inverted contrast, limited resolution, patient motion artifacts, and others, which makes it difficult to apply the already existing segmentation techniques for adult brain images. Therefore, the development of a robust technique to such problems is necessary. This project deals exactly with it, we propose a new structural method based on the learning of structural graphs with statistical attributes. (AU)