EFFICIENT NET-BASED DEEP LEARNING APPROACH FOR AUTOMATED BREAST CANCER HISTOPATHOLOGICAL CLASSIFICATION TOWARD PRECISION ONCOLOGY

Abstract

Breast cancer remains one of the leading causes of cancer-related mortality among women worldwide, highlighting the need for accurate and early diagnostic approaches. Histopathological image analysis plays a crucial role in breast cancer diagnosis; however, manual interpretation is time-consuming and prone to inter-observer variability. This study proposes an automated deep learning framework based on EfficientNet architectures for breast cancer histopathological image classification using the BreakHis dataset, comprising 7,909 microscopic images captured at multiple magnification levels (40X, 100X, 200X, and 400X). Transfer learning, data augmentation, and fine-tuning strategies were employed to enhance model generalization and address class imbalance. Two EfficientNet variants, EfficientNet-B0 and EfficientNet-B4, were evaluated for classification performance. Experimental results demonstrated that EfficientNet-B4 achieved superior predictive performance with a test accuracy of 97.3%, sensitivity of 97.6%, specificity of 96.6%, and test loss of 0.09, outperforming EfficientNet-B0, which obtained 92.4% accuracy and 0.21 loss. The confusion matrix analysis confirmed robust discrimination between benign and malignant tissue samples. The findings indicate that EfficientNet-based models provide reliable and efficient computational support for breast cancer diagnosis and may contribute to improved clinical decision-making in digital pathology workflows. Future studies will focus on multi-class tumor grading, explainable artificial intelligence techniques, and validation using real-world clinical datasets to enhance translational applicability in precision oncology.