An Improved Multi-Class Breast Cancer Classification and Abnormality Detection Based on Modified Deep Learning Neural Network Principles
Josephine D.C. Jullie, J. Sudhakar, Vidhya T. Helan, R. Anusuya, G. Ramkumar · 2024
Over the past decade, new lines of inquiry have been sparked by the pressing need for earlier identification of breast cancer. Breast cancer patients, according to the World Health Organization (WHO), have a better chance of survival if their illness is diagnosed at an early stage. Earlier illness diagnosis and treatment have the potential to reduce mortality rates in many regions of the world. Computer-aided diagnostic (CAD) techniques are routinely used to detect and diagnose a wide range of abnormalities. In recent years, the CAD system has become increasingly popular as a means to enhance precision across a variety of academic disciplines. The CAD systems’ outputs are accurate and need minimal human intervention. In this study, researchers employed a novel neural network method known as differential evolution to categorize breast cancer subtypes. The DEM method employed a Deep Neural Network (DNN) classification model. The feature vectors were generated using the EfficientNet feature extractor, and the parameters were fine-tuned with the help of Mayfly Optimization. The investigators needed a method to categorize hyper-spectral data, so they turned to deep learning. In this chapter, we used DNN to find hidden details in the data. The DNN is used at several stages of processing for breast cancer data categorization. The system was put through its paces using the UCI repository’s Wisconsin Breast Cancer Dataset (WBCD). Multiple train-test splits were performed on the dataset. The effectiveness of the networks is measured in terms of their accuracy, sensitivity, specificity, precision, and recall. The results showed a success rate of 98.72%, which is better than that of other state-of-the-art techniques. The results prove the method is far more effective than the one currently in use.