Intelligent Breast Cancer Diagnosis with Optimized Long-Short Memory Network Using Mammogram Images

Breast cancer is considered one of the most common cancers in women worldwide, with a high death rate. Still, one of the most effective methods for identifying breast cancer is medical imaging. Unfortunately, human experts require a lot of time to detect images. This work proposes classifying mammography images into three categories: normal, benign, and malignant, using the proposed Long Short Term Memory (LSTM) with Attention Mechanism (AM). The Mammographic Image Analysis Society (MIAS) dataset is used in this suggested effort for categorization. The fact that most of the now available datasets have imbalanced dataset issues-a sign that there are significantly more cancer patients in one group than the others-makes the problem even more challenging. Recognizing patterns in the imbalanced dataset has recently become a significant trust area in all domains. Choosing the tuned hyperparameters is essential when using LSTM models with AM. The Capuchin Search Word (CSW) algorithm determines the optimal hyperparameter values for the LSTM architecture. The experimental research indicated that the proposed LSTM with AM has 99.24% mean classification accuracy. This model, which has a more significant recognition rate and requires less computing time, will be the most effective option for recognizing breast cancer.