A Transformative Approach for Breast Cancer Detection Using Physics-Informed Neural Network and Surface Temperature Data

Abstract

Abstract Early detection is the most effective defense against breast cancer. Mammography is a well-established X-ray-based technique that is used for annual or biennial screening of women above age of 40. Since the dense breast tissue sometimes obscures the cancer in an X-ray image, about 10% of screened women are recalled and undergo additional adjunctive modalities, such as ultrasound, digital breast tomosynthesis, or magnetic resonance imaging. These modalities have drawbacks such as additional radiation dosage, overdiagnosis, and high cost. A new concurrent multispectral imaging approach was recently presented to eliminate the high recall rates by utilizing the breast surface temperature data with an inverse physics-informed neural network algorithm. This method utilizes the bioheat transfer modeling as the governing physics equations and conducted inverse heat transfer modeling using infrared temperatures to predict the presence of a tumor heat source. Validation of the predicted tumor size and location was conducted on a biopsy-proven breast cancer patient using infrared temperature data captured of the breast surface and pathology reports. A regression analysis between the predicted temperatures and infrared temperatures showed a coefficient of determination of 0.98. The absolute error in the predicted tumor size was 0.4 cm and the maximum absolute error in tumor location was 0.3 cm. The proposed approach shows promising results and performance. However, additional testing with more patients is required to quantify the standard deviation in the prediction and establish the sensitivity and specificity of the machine learning technique.