Leveraging EfficientNet-CNN for Accurate Diagnosis of Breast Cancer from Ultrasound Images

Abstract

Breast cancer is one of the most common malignancies in women worldwide. Because early detection is essential for effective treatment, researchers have investigated a number of methods to aid radiologists in the early detection of breast cancer. Convolutional neural networks (CNNs) are a promising method for diagnosing breast cancer when applied to the analysis of breast ultrasound images. This study investigates the classification of benign, malignant, and normal breast ultrasound images using CNN models with transfer learning, specifically the EfficientNetB0 architecture. 780 breast ultrasound images were extracted from the BUSI database. Although the EfficientNet architecture has fewer parameters than other CNN models, it provides faster and more accurate results, which is one of its main advantages. EfficientNetB0 architecture achieved 83.33 percent accuracy, demonstrating its ability to accurately classify breast ultrasound images.