Non-Invasive Breast Cancer Screening: Leveraging InceptionV3 and GoogLeNet for Thermal Image Analysis

Breast cancer remains a leading cause of mortality worldwide, highlighting the need for early detection. This research demonstrates the potential of combining thermal imaging and deep learning for non-invasive, cost-effective breast cancer screening. This study applies deep learning models, specifically InceptionV3 for feature extraction and classifiers such as GoogLeNet, DenseNet121, and MobileNetV2, to analyze thermal images for breast cancer diagnosis. Using the Database for Mastology Research (DMR), thermal images were preprocessed and augmented to improve the dataset. Transfer learning with pre-trained models enhanced diagnostic accuracy. The InceptionV3 + GoogLeNet model achieved the highest performance with an accuracy of 96.49%, an AUC of 99.72%, a precision of 98.25%, and an F1-score of 96.55%, outperforming other combinations. Despite the promising results, the study has limitations, including a relatively small dataset and the need for further validation. Nevertheless, it paves the way for improved early detection strategies, potentially reducing breast cancer mortality rates globally.