Breast Cancer Prediction in Mammogram Images using EfficientNet Based Hybrid Deep Learning Model

Scientists are interested in this area since many women with breast cancer do not seek medical attention until the disease has progressed. Genetics, hereditary factors, and inactivity are all contributors to the development of breast cancer. Breast cancer is death in women in India, and its incidence is rising across the country. Women over the age of 45 have an increased risk of contracting this illness. When compared to prose, pictures tell a thousand words. As computing power has increased, a number of methods have emerged for automatically gleaning previously undetectable details from photographs. Among the many fields where the processed photos have been put to use is medicine. The presence of breast lumps in the breast area is a necessary and sufficient factor in the progression of breast cancer in women. Therefore, a more trustworthy method is required to lessen the amount of pointless biopsies performed while diagnosing breast cancer. Doctors can make better judgements and reduce the number of needless biopsies with the use of computer-aided diagnostic technologies. For this purpose, the study applied the deep learning-based model for classification. Initially, three various steps such as normalization, contrast enhancement and noise removal during pre-processing. K-means clustering algorithm is used for segmentation, then the features are extracted by using pre-trained model called EfficientNet. The hybrid model called sparse auto-encoder based Long-short Term Memory (SAE-LSTM) is used. The proposed model efficiency is tested on mammographic image analysis society (MIAS) in terms of various metrics. The results indicated that the projected model is achieved a better performance than other counterparts.