Progressive Deep Learning Approach for Breast Cancer Diagnosis in Thermal Images

Worldwide, breast cancer is one of the leading causes of cancer-related deaths. However, death rates can be considerably lowered by detecting the disease early with imaging techniques like mammography and ultrasound. Medical thermal imaging is a useful non-invasive technique to spot anomalies. Early diagnosis and improved screening accuracy are being achieved through the application of machine learning-based risk prediction models. For the purpose of feature extraction and breast cancer classification, this work makes use of five cutting-edge convolutional neural network models as well as vision transformers. Annotated thermal images of both healthy and diseased breast tissues are included in the dataset. Features are extracted and concatenated for Independent Component Analysis (ICA). A Random Forest classifier is then trained using the updated features. This method highlights the advantages of integrating feature transformation techniques with CNN architectures to increase classification accuracy.