A Prognostic Framework for Post-Operative Patient Survival Prediction in IoMT

Abstract

The study presents an Internet of Medical Things (IoMT) framework designed to predict patient survival outcomes through the evaluation of a post-thoracic surgery scenario. We employ a multi-layered IoMT framework that integrates various sensors and medical devices for real-time data collection, efficient data transmission, and data analysis. Utilizing a set of eight traditional and ensemble machine learning classifiers, along with neural networks optimized using grid search, we establish a baseline performance for the framework's capability in predicting post-surgical survival rates. However, as individual machine learning classifiers exhibit suboptimal performance across the performance metrics used, we combine the individual strengths of these classifiers to construct a stacking approach. The stacked classifier which incorporates a multi-layer perceptron as the final estimator achieved significant results, including a high accuracy of 0.90, precision of 0.87, and recall of 0.93. These metrics not only indicate a high post-operative survival detection rate but also demonstrate a balance of low bias and high variance performance, ensuring that the model is both accurate and reliable in varying IoMT scenarios.