Metaheurísticas aplicadas à detecção, localização e classificação de tumores mamários compostos por múltiplos focos via termografia

Abstract

Being breast cancer the most incident neoplasia globally, recently auxiliary detection and monitoring techniques have been proposed, with thermography standing out due to its low cost. In this study, the use of meta-heuristics is proposed for the detection, localization, and classification of multiple-focus breast tumors through the use of skin surface temperature through an inverse analysis, simulating data obtained through thermography. Regarding the direct part of the proposed geometric inverse problem, the nonlinear Pennes equation was used in a 2D semicircular geometry representing the breast with multiple tissues and cancer foci. The solution to this equation is performed by the finite element method. For the inverse part of the problem, initially a comparative analysis between Genetic Algorithm, Differential Evolution, and an Self-adaptive Differential Evolution is conducted for the search of tumor geometric parameters using synthetic data simulated by the direct problem composed of two foci. The Self-adaptive Differential Evolution algorithm stands out for its automatic parameter adjustment capability and effectiveness in detecting two foci, presenting superior numerical results. Subsequently, the Self-adaptive Differential Evolution was applied to estimate multiple foci and classify tumors according to the system proposed by the Union for International Cancer Control, as it was the algorithm with the best performance. The algorithm succeeded in distinguishing between suspicious and cancerous breasts and classifying them according to the primary focus, considering the influence of external temperature. Keywords: Metaheuristics.