Deep Learning-Based Anomaly Detection of ECG Signals for Telemedicine Applications
N. S. Akhila, Sabeena Beevi K, Bejoy Abraham · International Journal of Engineering Trends and Technology · 2024
The growing global emphasis on healthcare and protection issues underscores the importance of point-of-care (POC) technologies. These technologies play a crucial role in delivering cost-effective solutions for telemedicine, further emphasizing their significance in the healthcare landscape. Although POC has clear benefits, it does not provide a primary patient diagnosis. Hence, this study focuses on the primary diagnosis of cardiac diseases by detecting abnormalities in the Electrocardiogram (ECG) on the patient's side itself. Achieving accurate diagnosis necessitates access to patients' confidential data. However, transmitting this sensitive information across a public network may lead to numerous security concerns. Additionally, serious privacy issues could arise since personal health information might be revealed to unauthorized individuals. This paper proposes a novel long, short-term based deep-learning technique for detecting the abnormality of ECG signals. The conveyed data's secrecy, security, and privacy are reinforced by employing a steganographic technique reliant on the Fast Walsh Hadamard transform. Features like spectral entropy and instantaneous frequency are used to increase the LSTM network's accuracy. Diverse transforms and their corresponding reconstruction methods are analysed, and it was observed that Walsh-Hadamard transforms are particularly well-suited for this specific application. This is primarily attributed to their compression capabilities, which effectively reduce the storage space needed for the data. The method being suggested is evaluated against traditional approaches outlined in the current literature.