Enhancing Breast Cancer Detection:Comparative Analysis of Convolutional Neural Network Architectures with Fine-Tuning
Sanchit Manwal, Gaurav Shankar Pandey, Navdeep Singh, Vihan Singh Bhakuni, Vikrant Sharma, Satvik Vats · 2024
A significant worldwide health problem that is more widespread among women in the modern era is breast cancer. Many methods have been investigated for identification of breast cancer problem, including Convolutional Neural Networks (CNN) and pre-trained models. Our study aims to enhance the efficiency and precision of breast cancer diagnosis by employing deep learning with transfer learning. In this research, we formulated different models that leverage deep learning techniques and transfer learning. The main aim was to improve and compare breast cancer detection accuracy of different architectures of CNN namely VGG16, INCEPTIONV3, XCEPTION and VGG19 by discerning between cancerous and non-cancerous regions in histopathology images. Through the utilization of a pre-trained model of architecture and subsequent fine-tuning, we achieved the best training accuracy of approximately 91.96% on the training dataset and 90.96% on the validation dataset using the VGG16 architecture.