Deep Learning (CNN) techniques for the Classification of Breast Cancer using Ultrasound Images A Review
Archana Singh, Surya Prakash Mishra, Prateek Singh · Journal of Emerging Technologies and Innovative Research · 2025
Breast Cancer, one of the main causes of death for women globally, highlights the vital need of an early and precise diagnosis. Ultrasonography, the first and primary diagnosis clinical investigating method lacks precise decision due to noise, shadow, contrast, and variations in tumor appearance. Convolutional Neural Networks (CNNs) in particular have shown remarkable achievements in medical image processing using deep learning models. In this paper, we have analyzed five popular pre-trained deep learning models— ResNet50, EfficientNetB0, VGG16, MobileNetV1 and DenseNet121 on publicly available dataset comprising 780 ultrasound images of both benign and malignant breast tumours from Baheya Hospital, Egypt. According to the results, the EfficientNetB0 network emerged as the best model for determining whether breast cancer is benign or malignant. In the study conducted, the Efficient-Net B0 model demonstrated impressive results with an accuracy rate of 96.80% among all the models with considerably less training time and minimal parameters. The ability of the transfer learning from Imagenets database module has also remarkably improved the accuracy of breast tumour categorization using ultrasound images.