Pre-processing of Cardiac Magnetic Resonance Images for Myocardial Segmentation

• Grant number: 22/10409-2
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: February 01, 2023
• End date: June 30, 2024
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Principal Investigator: Regina Célia Coelho
• Grantee: Sophia Silvestre Camargo
• Host Institution: Instituto de Ciência e Tecnologia (ICT). Universidade Federal de São Paulo (UNIFESP). Campus São José dos Campos. São José dos Campos , SP, Brazil

Abstract

Imaging diagnosis is an efficient method in medicine. Cardiac magnetic resonance imaging (CMR) allows the assessment of anatomy and possible pathologies. In addition, applying deep learning (DL) contributes to the examination evaluation process, reducing the physician's work, which allows him to have greater availability for patient care.The first step is to generate a set of quality images for the DL model to learn the desired characteristics of the images correctly. Therefore, the CMR images must have adequate brightness and contrast to identify the contours and edges of the different regions in each section. For example, on examination, the chest appears almost entirely in the images; reducing the amount of information by limiting itself to the heart region contributes to neural network training. This work aims to apply techniques to improve the quality of images, adjusting brightness and contrast to facilitate the identification of the contours and edges of the left ventricle and myocardium. In addition, identifying the regions of interest and eliminating non-relevant information will facilitate training the U-Net convolutional neural network to be used in the segmentation.