Breast Cancer Detection using Histopathological Images

Breast cancer remains a major global health problem, and accurate and timely diagnosis of the disease is critical for improved survival rates. While traditional diagnosis methods like mammography and biopsy are well established, they are usually plagued by high false-positive rates and failure to discriminate between subtle histopathological patterns. Deep learning has made significant progress in medical image classification over the last few years, and Convolutional Neural Networks (CNNs) like ResNet and DenseNet have been demonstrating promising results. However, single models are not without their weaknesses, including vanishing gradients in deeper networks and poor feature propagation. To address these challenges, we introduce an ensemble model of DenseNet and ResNet architectures for breast cancer classification using histopathological images of the BreaKHis dataset. DenseNet offers improved feature propagation through densely connected layers, while ResNet solves the degradation issue via residual connections. Our ensemble model leverages the complementary strengths of the two architectures to offer improved feature extraction and classification accuracy. Experimental results show that the proposed ensemble model achieved an accuracy of 97.5%, outperforming the performance of the single DenseNet model (96.3%) and the single ResNet model (86.1%). This study demonstrates the potential of deep learning ensembles in improving the accuracy of breast cancer diagnosis and reducing misclassification in histopathological examination.