Number of Convolution Layers and Convolution Kernel Determination and Validation for Multilayer Convolutional Neural Network: Case Study in Breast Lesion Screening of Mammographic Images

Mammography is a low-dose X-ray imaging technique that can detect breast tumors, cysts, and calcifications, which aid in detecting potential cancer the early stage reduce mortality rate. This study employed multilayer convolutional neural network (MCNN) to screen lesions with mammographic images. Within region of interest, specific bounding box used extract feature maps before automatic image segmentation classification are conducted. These include three classes, namely, normal, benign tumor, malignant tumor. Multiconvolution processes kernel convolution operations have noise removal sharpening effects better than other processing methods, strengthen features desired object contour increase classifier’s accuracy. However, excessive layers will computational complexity, time, training time for classifier. Thus, this aimed determine suitable number kernels achieve classifier high learning performance accuracy, case lesion screening The Mammographic Image Analysis Society Digital Mammogram Database (United Kingdom National Breast Screening Program) was experimental tests kernels. optimal evaluated using accuracy (%), precision recall F1 score test validate most MCNN model architecture.