DAIIP: Distributed Artificial Intelligence-based IoMT Platform for Clinical Data Collection and Applications

• Published in: 2024 5th International Conference on Computer, Big Data and Artificial Intelligence (ICCBD+AI) pp. 454 - 465
• Main Authors: Fang, Keyu, Qin, Xiaoli, Wang, Junjun, Zhang, Xin, Chen, Ping
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.11.2024
• Subjects: Artificial intelligence; Deep learning; Distributed AI; Distributed databases; Edge computing; Hospitals; Internet of Medical Things; Optical character recognition; Physiology; Predictive models; Time series analysis; Time-series Prediction
• DOI: 10.1109/ICCBD-AI65562.2024.00082

The massive onsite time-series clinical data produced by Internet of Medical Things (IoMT) can bring valuable information and immense potentials, thus enpowering a new wave of applications such as clinical decision support. However, current IoMT platforms is limited to data collection, some of them can cause data silos, further leading to data loss. To address these challenges, we present a Distributed Artificial Intelligence (DAI)-based IoMT Platform for Clinical Data Collection and Applications, integrating time-series clinical data streaming processing with deep learning (DL)-based applications. This study focuses on optimizing the OCR model to improve the accuracy and processing speed of medical equipment data, thereby enhancing data collection compatibility. Additionally, we have integrated the MEWS scoring system with five fundamental physiological parameters and optimized the Autoformer time series prediction model. These enhancements enable the platform to intuitively depict changes in patients' physiological states. We validated the feasibility and effectiveness of our proposed platform by constructing an experimental setup that simulates real hospital scenarios and conducting comprehensive tests. The experimental results demonstrate that our platform effectively processes clinical data and supports clinical decision-making. These findings underscore the platform's potential and value in the IoMT field.