Hybrid Transfer Learning of Mammogram Images for Screening of Micro-Calcifications
Shanker MC, M Vadivel · International Journal of Electrical and Electronics Engineering · 2022
Breast cancer is a deadly disease occurred in women due to modern lifestyles. It is a second serious disease to women's health, both physically and psychologically. Computer-aided design-based approaches support clinicians in the early diagnosis of breast cancers. It uses machine learning and deep learning algorithms to detect breast cancers in mammogram images accurately. This work proposes the hybrid ResNet and Bidirectional Long Short-Term Memories (BiLSTM) based transfer learning model for automatic micro calcifications classification in mammogram images. The hyperparameters of the proposed hybrid model are tuned by the Eurasian oystercatcher optimizer (EOO) to get superior performance in classification. The proposed model is applied in the MIAS database for classifying benign and malignant stages and compared against previously proposed models. The proposed model achieved a better result in all iterations for accuracy, precision, recall, and specificity 98.4%, 98.2%, 98.4 % and 99.4 %, respectively.