Deep Learning based Breast Cancer Classification on Thermogram

Breast cancer disease is a global matter of concern for women all over the world. There are now various medical imaging approaches available for breast cancer diagnosis, including mammography, magnetic resonance imaging, ultrasound, and digital thermal imaging, which is presently finding its way into this domain in combination with mammography. Infrared imaging has been proven to be less harmful than mammography and can provide findings several years before mammography can identify cancer. The objective of this work is to combine thermography with deep learning to offer new insight into how to develop better breast cancer prediction. The methods and techniques employed in this study are based on a deep learning model that has been pre-trained to predict breast cancer using thermal images. With 100 patient thermograms (FL-R different view angles), the proposed methodology revealed a considerable improvement in accuracy to 92.66 % utilizing the GoogleNet model, which performed well in comparison to the AlexNet model for 20 epochs. As a result, we can conclude that the suggested deep learning model is extremely successful in breast cancer prediction.