IoT-Based Hybrid Fuzzy LSTM-RNN for Secure Disease Prediction in Healthcare EHRs

The integration of Fuzzy Logic and Long Short-Term Memory Recurrent Neural Networks (LSTM-RNN) is employed to handle healthcare data, leading to a significant improvement in the prediction of unknown disease outcomes and notably enhancing reliability and accuracy. In this research, we propose an integrated IoT-based healthcare data management system with Fuzzy Long Short-Term Memory Recurrent Neural Network (IF-LSTM-RNN) for disease prediction and diagnosis. Our approach includes gathering data via IoT devices, preprocessing through min-max normalization, and utilizing IF-LSTM-RNN for predictions. Clinical data is first collected and preprocessed, from which the health outcomes of patients are then predicted through IF-LSTM-RNN. The anticipated data is securely stored in Electronic Health Record (EHR) systems, making it more secure and providing accurate predictions. To evaluate the performance of the proposed system, we applied it to a dataset comprising glucose concentrations from 12,612 data points of five monitored subjects with diabetes. The IF-LSTM-RNN outperformed traditional techniques (Random Forest, Support Vector Machine, and K-Nearest Neighbors) with an accuracy of 99.62%, precision of 98.71%, recall of 97.91%, an F1-score of 98.64%, sensitivity of 98.95%, and specificity of 97.88%. The IF-LSTM-RNN also achieved a correct classification rate of 99.37% with an execution time of approximately 1.28 seconds. The results demonstrate that the proposed framework offers a viable solution for secure and effective healthcare data management and prediction in IoT environments.