Hybrid Deep Learning Model for Enhanced Breast Cancer Diagnosis Using Histopathological Images
Subrahmanyasarma Chitta, Shubham Sharma, Vinay Kumar Yandrapalli · Procedia Computer Science · 2025
Breast cancer is a significant global health concern, necessitating early and precise diagnosis for effective treatment. Traditional diagnostic methods, while reliable, are often labor-intensive and prone to human error. This study investigates advanced deep learning (DL) models to enhance the precision and efficiency of breast cancer detection using histopathological images. Although standalone DL models, such as convolutional neural networks (CNNs) and vision transformers (ViTs), have demonstrated substantial promise, they often face challenges in capturing the complex features of medical images. To address these limitations, we propose a novel hybrid architecture integrating a customized EfficientNetV2 with a modified ViT model. The EfficientNetV2 is optimized for robust feature extraction, while the ViT is adapted to model spatial relationships effectively. The hybrid framework merges features from both models, processing them through a dense layer to achieve improved Classification accuracy. Utilizing the BreaKHis dataset, the hybrid model achieved exceptional results, including an AUC of 0.91, an accuracy of 99.56%, precision of 99.56%, recall of 99.55%, and an F1-score of 99.56%. These findings represent a significant improvement over conventional approaches. By integrating features from convolutional and transformer-based models, this hybrid architecture showcases potential for broader applications in medical image analysis. Future research directions include incorporating multi-modal data and exploring unsupervised learning techniques to enhance robustness and generalization across diverse datasets.