STOA based Feature Selection with Improved LSTM Model for Breast Cancer Diagnosis in IoT

Medical and health care have benefited greatly from IoT advancements. This technology helps both patients and doctors get a clear picture of a wide range of illnesses and make accurate diagnosis. The problem of low diagnostic accuracy in breast cancer diagnosis is, however, already included in the standard research approaches. Maintaining a strong foundation for breast cancer management and therapeutic advancement, early detection is essential. However, due to the nonappearance of indications in the early stages, early identification of cancer is challenging. As a result, cancer is still one area of medicine that scientists are working to advance in terms of detection, prevention, and therapy. The use of deep learning methods in mammogram processing has helped radiologists save money in recent years. In the current breast mass classification methods, deep learning knowledges like a (CNN). Although CNN-based systems have improved upon the pictures, several problems remain. Ignorance of semantic characteristics, analysis bound to the present patch of pictures, missing patches in low-contrast mammograms, and ambiguity in segmentation are all problems that need to be addressed. Because of these problems, this study's primary impartial is to create a deep learning-based system for classifying breast tumours in mammographic images as malignant or benign utilising two approaches: feature selection and classification. In this study, a recurrent neural network is employed for classification after the unnecessary data has been removed using the Sooty Tern Optimization Algorithm (STOA). Elite opposition-based learning optimally selects the weight and bias of Long-Short Term Memory (LSTM) (EOBL). Furthermore, two publicly accessible datasets of mammographic pictures are used to equivalence the projected approach to preexisting categorization systems. Comparative studies showed that the suggested strategy outperformed previously developed mammography categorization algorithms.