Iris Project: computer vision system for assessing human body composition

• Grant number: 20/05792-6
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Start date: June 01, 2021
• End date: April 30, 2022
• Field of knowledge: Interdisciplinary Subjects
• Principal Investigator: Laísa Ribeiro Silva de Abreu
• Grantee: Laísa Ribeiro Silva de Abreu
• Company: ACRM Inovações Tecnológicas Ltda
• CNAE: Atividades de atenção à saúde humana não especificadas anteriormente
• City: Indaiatuba
• Pesquisadores principais: Mateus Modesto
• Associated researchers: Bruna Bittar Martins ; Marco Túlio Rodrigues Marques
• Associated scholarship(s): 21/06000-9 - Iris Project: Computer vision system for evaluating the human body composition., BP.TT; 21/08872-3 - Development of a visual analysis composition software based on computer vision., BP.TT; 21/05889-2 - Body composition assessment by anthropometry and bioelectrical impedance in software development context for visual analysis of human composition., BP.TT; 21/03786-1 - Iris Project: computer vision system for assessing human body composition, BP.PIPE

Abstract

The analysis of body composition refers to the measurement and qualitative and quantitative characterization of the different types of tissues that make up the human body, contributing to the determination of the nutritional status of the individual and populations. Its importance lies in the widely documented association of these parameters with a wide spectrum of diseases such as cardiovascular diseases, diabetes, cancer, metabolic syndrome, osteoporosis and osteoarthritis. In addition to the composition, the body distribution of adipose tissue has been raised as a factor of high impact on cardiometabolic risk. The traditional methods used in estimating human body composition include anthropometric measurements obtained with manual use of adipometers, densitometry - the Dual Energy X-Ray Absormetry (DXA), bioelectrical impedance, computed tomography, among others, whose applications are scientifically limited and / or high cost. Thus, given the obesity epidemic and its chronic comorbidities, scientific and economic interest in the accurate and low-cost establishment of body composition and distribution has increased over the past few years. Thus, seeking a solution for universal use, with greater precision and reduced cost is desirable and opens the way for several applications. The use of computer vision in the development of specialized systems has received great academic contributions that allowed it to be one of the bases for solutions in the Fourth Industrial Revolution (industry 4.0, health 4.0, smart cities, smart farms, etc.). Computer vision solutions have been intensified, mainly, with the use of deep neural networks, which can be trained for more assertive assessments also in the context of nutrition. Thus, the objective is to apply their bases in the health field and develop neural network algorithms that are accurate enough to be able to measure the individual's body composition through images, showing the percentage of body fat and, if possible, its distribution. (AU)