Estimation of Breast Tumor Parameters by Random Forest Method with the Help of Temperature Data on the Surface of the Numerical Breast Model

Abstract

The second most frequent reason for cancer-related fatalities in women is breast cancer. When a condition is identified early, better treatment choices are available. Different temperature patterns are seen on the breast surface due to the tumors, which change blood perfusion rate and metabolic heat production. Thermography is an infrared imaging technology for breast cancer screening that records temperature variations. The temperature dataset on the surface of the breast that corresponded to the tumor’s diameter and the location was needed for the current study, but such actual data are not accessible. Thus, the modeling and development of a dataset constitute the initial component of the current study. The bio-heat transport equation is solved using COMSOL multiphysics software, and the model consists of a spherical tumor inside of a hemispherical breast model. By changing the sizes and positions of the tumor inside the breast during simulations, a reliable dataset is created. The training and testing of the dataset produced from the simulations using the random forest method make up the second portion of the current study. Breast skin temperature is used as an input in a random forest machine learning algorithm in the current work to determine the diameter and location of the tumor inside the breast. The diameter and area of the tumor location are estimated by a trained random forest algorithm. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.