Patient Monitoring based on ICU Records using Hybrid TCN-LSTM Model

Patient monitoring system is very important as it facilitates the constant and effective monitoring of patient's health with a view of enhancing their result. However, conventional monitoring systems have problems concerning scalability, real-time analysis, and processing of difficult time series data. Based on MIMIC-III dataset, which is a widely used dataset with intensive EHR records of ICU patients, this paper designs a cloud-based architecture utilizing TCN and LSTM models for highly usable health monitoring with considerably less time complexity. The proposed TCN-LSTM approach is used to incorporate the high temporal resolution of TCN to capture the long-range temporal dependency and the ability of LSTM for sequential processing and decision making to develop an effective analysis of the patient data. Based upon a cloud computing structure, this system, which is integrated into a data processing system, also provides reliable storage and access to data as well as fast processing time. According to results given on MIMIC-III dataset we can improve the inaccuracy reading patient conditions to alert early enough or even personalised care. All the results point out the beneficial application of cloud-based TCN-LSTM architectures for improving patient health monitoring as a scalable and real-time platform with high predictive accuracy.