A Convolutional Network Approach for Breast Tumor Classification Using Ultrasound Images

Breast Tumor is still considered to be the leading disease that threatens the lives of women globally. However, the key to successful treatment and increased chance of survival is early detection. This study suggests a novel Deep Learning model designed for detecting breast tumors from the ultrasound images of breasts. Our approach uses features from the InceptionResNet Version 2 model for enhanced feature extraction as well as Gray Level Co-occurrence Matrix for higher texture analysis. The proposed system is assessed using the Breast Ultrasound Image dataset and achieved overall classification accuracy of 98.08%. The model not only contributes to the development of the current field of medical imaging but also offers a solution to increase diagnostic accuracies for the cases in different clinical settings.