On the detection of breast tumors using dynamic thermography and Self-adaptive Differential Evolution

Resumo

Recently, the clinical diagnosis of breast tumors using thermal analysis has gained increasing attention as a non-invasive and cost-effective alternative to traditional techniques. To enhance breast cancer detection and reduce false positive and false negative rates, several authors have suggested the use of dynamic infrared thermography. In this study, we propose a dynamic thermography approach for the detection of breast cancer with tumor focus. We model a circular tumor embedded within a multilayered breast structure composed of five distinct tissue layers. Based on simulated skin surface temperature measurements, the geometric parameters of the tumors are estimated using metaheuristic optimization techniques. To simulate the effects of real-world measurement conditions, random noise is added to the synthetic temperature data, where the noise is consistent with the magnitude typically observed in modern infrared cameras. The forward problem is modeled using a two-dimensional transient bioheat transfer model governed by Pennes’ equation, and it is efficiently solved via the Finite Element Method implemented with FEniCS. The inverse problem is formulated as an optimization task and tackled using Self adaptive Differential Evolution, an auto-adaptive variant of Differential Evolution. Numerical experiments, conducted both with and without noise, demonstrate that the proposed approach is a promising alternative for solving inverse problems in breast tumor detection through thermographic analysis.