DEEP LEARNING BASED BREAST CANCER IDENTIFICATION THROUGH ANALYSIS OF HISTOPATHOLOGICAL IMAGES

Abstract

Breast cancer is one of the most prevalent malignancies among women worldwide and remains a leading cause of cancer-related mortality. Early and accurate diagnosis is essential for improving patient survival rates and guiding appropriate treatment strategies. Histopathological image analysis is widely regarded as the most reliable diagnostic approach for breast cancer; however, manual examination of biopsy samples is time-consuming, labor-intensive, and prone to inter-observer variability among pathologists. These limitations highlight the need for automated and reliable diagnostic solutions.

To address these challenges, this study proposes an automated breast cancer detection framework based on deep learning and transfer learning techniques. The proposed approach employs the EfficientNetB0 convolutional neural network due to its ability to achieve high classification accuracy while maintaining reduced computational complexity. A pre-trained EfficientNetB0 model is fine-tuned using histopathological image patches from the Breast Histopathology Images (IDC) dataset to perform binary classification between invasive ductal carcinoma (IDC) positive and IDC negative samples.

Experimental results demonstrate that the proposed model achieves a training accuracy of 94.49% and a test accuracy of 94.46%, indicating strong generalization performance. Confusion matrix analysis reveals a true negative rate of 98.0% and a true positive rate of 74.5%, highlighting the model’s effectiveness in correctly identifying non-cancerous tissue while maintaining reliable cancer detection. These results suggest that EfficientNetB0-based deep learning models can serve as effective computer-aided diagnostic tools to support pathologists in breast cancer screening and clinical decision-making.