News

AI and IoMT Cloud Integration Yields Promising Gains in Remote Cardiac Monitoring

• By John K. Waters
• 08/01/2025

A new study has introduced a cloud-integrated, artificial intelligence-driven framework for remote patient monitoring, offering a potential advance in early detection and management of heart disease. The system combines Internet of Medical Things (IoMT) sensors, cloud computing, and transformer-based AI models to analyze patient data in real time, with notable improvements in classification accuracy and efficiency.

Researchers developed a Transformer-based Self-Attention Model (TL-SAM) that processes both spectral and spatial features of physiological signals, such as heart rate and blood pressure. The model employs a dual-branch structure—SpecSAM for spectral data and SpatSAM for spatial input—enhancing diagnostic accuracy by separately analyzing time-based and positional patterns in health data. Feature fusion from both branches is then used to classify cardiac conditions.

To improve the model’s performance, investigators applied an optimization technique called the Improved Wild Horse Optimization with Levy Flight Algorithm (IWHOLFA). This evolutionary computation method was used to fine-tune model parameters and improve learning convergence.

The study utilized the publicly available UCI Heart Failure Clinical Records dataset, comprising 299 patient records. After preprocessing and normalization, the model was trained and tested using standard evaluation metrics. TL-SAM achieved an accuracy rate of 98.62%, with a precision of 97%, recall of 98%, and an F1-score of 97%. These figures surpassed the performance of existing deep learning models, including CNNs, RNNs, and Vision Transformers.

In addition to outperforming traditional architectures, the model’s statistical superiority was confirmed through paired t-tests comparing TL-SAM to baseline methods. Researchers found significant differences in both accuracy and F1-scores (p < 0.05), indicating the model’s reliability in practical applications.

The system includes an alert mechanism that evaluates vital sign abnormalities and scores patient conditions in real time. Alerts are triggered when patient scores surpass established thresholds, enabling clinical personnel to respond swiftly to potential emergencies.

The TL-SAM framework utilizes wireless sensor networks and cloud infrastructure to enable continuous, non-invasive monitoring for heart disease. By transmitting and analyzing data remotely, the system minimizes the need for in-person diagnostics, which is especially advantageous in resource-limited or rural areas.

While the study focused on heart failure prediction, the framework is adaptable for other chronic diseases. Future work aims to expand the system’s capabilities through longer-term data collection and integration with broader IoMT ecosystems, further enhancing its predictive scope and scalability in real-world healthcare settings.