Fully Interpretable Deep Learning Model Using IR Thermal Images for Possible Breast Cancer Cases.

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Breast cancer is a major global health concern necessitating improved early detection methods.
• Thermography presents a promising, non-invasive, and cost-effective adjunctive tool for breast cancer screening.

Purpose of the Study:

• To explore the efficacy of machine learning, specifically Bayesian networks and convolutional neural networks (CNNs), for early-stage breast cancer diagnosis.
• To integrate Explainable Artificial Intelligence (XAI) for enhanced interpretability of diagnostic models.

Main Methods:

• Development of two diagnostic expert models (Model A and Model B) integrating thermal imaging and medical records.
• Model A utilized XAI with thermal images and medical data.
• Model B incorporated XAI, thermal images, medical data, and CNN predictions.

Main Results:

• Model A achieved an accuracy of 84.07% in breast cancer diagnosis.
• Model B demonstrated a higher accuracy of 90.93% by including CNN predictions.
• The integration of XAI significantly improved the interpretability and accuracy of the diagnostic models.

Conclusions:

• Explainable AI, combined with advanced machine learning techniques like Bayesian networks and CNNs, shows significant potential for improving early breast cancer diagnosis.
• The developed models offer a highly accurate and interpretable approach to breast cancer detection, supporting clinical decision-making.