Skin lesion analysis using multispectral images captured by a low-cost device

• Grant number: 22/11762-8
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: November 01, 2022
• End date: June 30, 2024
• Field of knowledge: Engineering - Electrical Engineering
• Principal Investigator: Leticia Rittner
• Grantee: Nathan Shen Baldon
• Host Institution: Faculdade de Engenharia Elétrica e de Computação (FEEC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil

Abstract

The analysis and follow-up of skin lesions are important for the diagnosis and treatment of some diseases, such as lupus. However, this task can be laborious, subjective, and limited when done only visually. In this case, spectral imaging combined with computational methods proved interesting. From this type of image, additional information beyond the visible spectrum can be obtained, besides the spectral signatures of the different regions of a sample, the latter being a characteristic curve of a material that can be used to identify it. Multispectral cameras, however, are expensive and infeasible to public hospitals (SUS, for example). It becomes interesting, then, to evaluate whether a low-cost multispectral camera would work for skin lesion analysis. Thus, this project seeks to build a low-cost multispectral camera from a Raspberry Pi board and a Raspberry Pi camera module and test it first on a simpler problem, like the analysis of agricultural products' quality, and then in the analysis of skin lesions. In addition, it is desired to propose an automatic method of analysis of these images, through image processing algorithms and machine learning. To achieve that, it'll be necessary: a Raspberry Pi board, a Raspberry Pi camera module, band-pass filters or LEDs of different wavelengths and a dark box with built-in lighting. Initially, the camera will be adapted and tested. After that, the multispectral images will be collected. Then, for the image processing and analysis, the following steps will be followed: noise removal, calibration, feature extraction (spectral signature being one of them) and the training of a classifier. As a result of this project, it is expected an objective assessment of the feasibility of the low-cost multispectral camera for acquisition of multispectral images of skin lesions. Also, a proposal of a method for automatic processing and analysis of these lesions will be presented.