AI-Driven Prediction of Chronic Kidney Disease Using Cloud-Based Health Records

Abstract

Any condition that occurs on a regular basis and it has no symptoms is suspicious enough of Chronic Kidney Disease (CKD), which goes undetected during a tender stage and has to incur irreparable damage to the kidney and cause surmountable healthcare expenses. These methods of diagnosis are more of a one off because there is no continuity and timing is essential in the early intervention process and diagnosis requires a period before continuity and exhibiting an iterative process can lead to its acquisition. This paper introduces a predictive model of cloud-based electronic health records (EHR)-driven AI-assisted early detection of CKD. The system combines the capabilities of more extreme machine learning and deep learning systems, including the Random Forest, the Support Vector machine (SVM), Deep Neural Networks (DNN), and a simplified hybrid RF-LSTM model, which unites the explainability of the Random Forest and the memory ability of long short-term memory networks that are (LSTM). To improve the performance of models, data preprocessing included data normalization, filling blank data and feature selection. Experimental outcomes indicate a superior performance of the hybrid RF-LSTM in terms of accuracy (0.96), Precision (0.95), Recall (0.96), F-1-score (0.95), and ROC-AUC (0.97) over the lone models. Implemented on Amazon Web Services (AWS) with the help of S3 (storing data), SageMaker (training), and Lambda (real-time execution) the system delivered its predictions within less than 300 milliseconds. Indeed, the results highlighted the possibility of integrating AI and cloud computing to provide comparatively inexpensive and scaled real-time CKD risk prediction approach, which can facilitate the provision of better patient outcomes and preventive care.

Ähnliche Arbeiten