Intelligent healthcare monitoring systems are emerging as a result of the proliferation of portable medical devices with Internet of Things (IoT) capabilities. Deep learning (DL) and the IoT help the healthcare sector by facilitating telemedicine, which lowers disease. It is imperative to identify and categorize arrhythmia stages early to diagnose the disease, since heart arrhythmia is a fatal condition that puts many lives at risk. Due to its poor accuracy, arrhythmia identification and classification remain very difficult in the medical field. A unique hybrid DL-based Internet of Things-enabled health surveillance system is suggested in this regard, offering better accuracy and fewer error rates in the prediction and classification of arrhythmias. Initially, the input Electrocardiogram data collected from the databases are subjected to a pre-processing stage in which irrelevant noise components are suppressed. Secondly, the useful critical features needed for disease prediction are selected using a Hybrid Leopard Seal Circle-Inspired Optimization approach. Next, a Hybrid Enhanced Spiking Sheaf Attention Neural Network is employed to predict and categorize arrhythmias according to the selected critical features. The Hybrid Enhanced Spiking Sheaf Attention Neural Network model's hyperparameters are optimized through the use of Artificial Rabbits Optimization. Lastly, the effectiveness of the proposed model is assessed using two publically accessible datasets: the Atrial Fibrillation Database (AFDB) and the MIT-BIH Arrhythmia Database. Experimental results show that the stated model achieves 99.4% accuracy on the AFDB and 99.2% accuracy on the MIT-BIH Arrhythmia Database.