Wavelet Pooling in Convolutional Neural Networks for Breast Cancer Detection with Ultrasound Images

Analyzing ultrasound breast cancer images is crucial for the early detection and diagnosis of breast cancer. It helps identify abnormal masses or lesions, especially in women with dense breast tissue. Early detection is important for successful treatment. Careful examination of ultrasound images can help identify size, shape, texture, density, and location of any identified masses. It plays a critical role in saving lives and improving outcomes for women with breast cancer. The primary objective of this study is to introduce a new approach for enhancing the efficiency of a standard convolutional neural network (CNN) for analyzing ultrasound breast cancer images. The idea suggested is to incorporate wavelet transformation as a substitute for the pooling process in CNN and the new architecture is named ‘The MobileNetV2 plus’. The MobileNetV2 plus architecture was developed. Through various measurement tools, its effectiveness was tested and confirmed, showing promising results with a high level of accuracy and satisfactory performance.