Hybrid Deep Convolutional Neural Network Approach for Detecting Breast Cancer in Mammography Images

Breast cancer is among the top causes of fatalities related to in females. Radiologists commonly use mammogram images detect breast tumors their early stages. However, mammography can produce low-contrast images, making it difficult and time-consuming segment abnormal regions. Deep convolutional neural networks (CNNs) are used for image evaluations. This study deep CNN models develop a computer-aided diagnostic (CAD) system feature extraction classification. The proposed approach consists three phases. In first phase, shallow, model comprising five layers, max-pooling one batch normalization layer, dropout layer was developed extract recombined novel features. second Inception-v3 label smoothing classification due its multiple filters with different sizes. third features were extracted using models. Infallible Euclidean distance-based nonlinear dimensionality reduction minimize dimensionality. Finally, Gini-index-based C4.5 decision tree binary from Digital Database Screening Mammography (DDSM) + Curated Imaging Subset DDSM (CBIS-DDSM) Mammographic Image Analysis Society (MIAS) datasets. hybrid achieved 99.52% accuracy, 96% AUC on CBISDDSM dataset, an accuracy 97.53% value 97% MIAS dataset. Compared other cutting-edge CAD systems, higher by combining in-depth across both