SECURE AND EFFICIENT CLASSIFICATION OF CHEST X-RAY IMAGES USING CRYPTOGRAPHY AND MACHINE LEARNING TECHNIQUES
P L Lekshmy, Abdul Rahiman · International Journal for Multiscale Computational Engineering · 2024
Chest X-ray images are an essential diagnostic tool in the medical field, but transmitting and storing these images can raise security concerns. Privacy-preserving techniques in healthcare databases protect sensitive patient information from unauthorized access, breaches, and misuse. Using a combination of elliptic curve cryptography (ECC) for encryption, a convolutional neural network (CNN) for feature extraction, and a long short-term memory (LSTM) network for classification, the aim of this research is to develop a safe and effective classification system for chest X-ray images. The research integrates hyperparameter optimization, essential for LSTM, as it dramatically impacts the model's performance. Additionally, the research employs an oppositionbased learning strategy, the traditional wild horse optimization (WHO) approach and an enhanced WHO (EWHO), to enhance the classification performance. The research objectives are to diagnose and classify chest X-ray images into four types: normal lung, COVID-19-infected lung, lung opacity, and viral pneumonia. The proposed opposition-based learning model improves the accuracy, efficiency, and consistency of diagnostic and treatment decisions. The proposed EWHO-configured LSTM approach attains 97.4% accuracy for an efficient classification system for chest X-ray images. The research highlights the importance of privacy-preserving techniques in healthcare databases to protect sensitive patient information from unauthorized access, breaches, and misuse.