CTF-ANET: Clinical Time Frequency Aware-Attention Network for cardiovascular disease classification in wearable IoT network

Cardiovascular diseases are the primary causes of illness and deaths worldwide. The continuous monitoring and earlier identification of cardiovascular disease are essential for prevention. An advancement of wearable Internet of Things (IoT) devices equipped with sensors, like an electrocardiogram (ECG) has revolutionized health monitoring by facilitating the collection of data outside medical settings. However, existing approaches for cardiovascular disease classification frequently struggle with the dynamic nature and complexity of clinical databases, demanding the development of more refined techniques. In this research, the Clinical Time Frequency Aware-Attention Network (CTF-ANET) is introduced for cardiovascular disease classification in a wearable IoT network. Here, a wearable IoT network simulation is performed initially, and thereafter, an ECG signal is acquired from a wearable IoT sensor. After that, anomaly signal detection is carried out utilizing Outlier Detection based on Neighbor Difference (ODND) algorithm, and then, the anomaly signal is removed. Afterwards, clinical and joint-time features are extracted, and next, transformation-based augmentation of feature components is accomplished. Lastly, cardiovascular disease classification is done employing CTF-ANET. Here, the Grunwald-Letnikov derivative-based learning rule is proposed to train CTF-ANET. Additionally, CTF-ANET has attained a maximal accuracy of 96.641%, sensitivity of 95.675% and specificity of 96.966%.