Breast Cancer Detection Using SEResNeXt and Swin Transformer
Sashank Talakola, Devi Vara Prasad Pattupogula, Raja Karthik Vobugari, Balaji Ummadisetty · 2023
In this work, we intend to device a methodology to solve the issue of Medical Image Classification on cancer patients using Deep Learning techniques. According to the World World Health Organization, the most common occurring cancer globally is breast cancer. In the year 2020, 2.3 million new breast cancer cases have been diagnosed. Due to this nearly 685,000 individuals died. From the 1980s regular checkup in mammography screening had been proposed by the health authorities which resulted in a 40 % reduction in several countries. At the moment, beginning stages of breast cancer can be detected using mammography specialized radiologists. In the past few years it has been reported that, there is a shortage among radiologists in several nations. This process could be automated using Deep Learning techniques, which would reduce the number of human-made errors and amount of time taken to inspect the MRI scan. Henceforth, in the work using feature vectors from SEResNeXt and applying Attention mechanism over these vectors to efficiently and accurately classify the MRI scans and we have obtained a probabilistic F1 score of 0.61 and an accuracy of 0.96.