KDAIDL: Kidney Disease Detection Using Artificial Intelligence and Deep Learning

A remarkable increase in chronic kidney disease (CKD) is in public health nowadays. Timely diagnosis is crucial for the proper treatment of this illness. This work introduces a deep learning model capable of properly predicting the early stages of the disease. The evaluated models are: Deep Belief Network (DBN), Gated Recurrent Unit (GRU), and the proposed innovative method, CNN+DBN model. Our study compares three models under two conditions: without hyperparameter tuning and with hyperparameter adjustment. After doing hyperparameter tuning on a chronic kidney disease dataset, the model attained an accuracy of 99.2%. This accuracy is significantly higher accuracy as compared to other researchers. The DBN model demonstrates superior performance compared to other models regarding accuracy, recall, F1 score, AUC, and precision. This study introduces a CNN-based model that integrates the deep learning capabilities of the DBN with the feature extraction strengths of the CNN. It effectively captures spatial dependencies in the collected data. The extracted features enable the model to process and predict the early stages of kidney disease. The combination of the CNN and DBN models demonstrated high accuracy. Hyperparameter tuning improves accuracy to 98.15%, exceeding the performance of GRU+DBN models. The CNN+DBN model processes this data further. The hyperparameter tuning and the combined model attained an accuracy of 98.15%, surpassing the performance of the GRU+DBN models. The interpretability and performance of deep learning models are critical factors in the development of efficient models. The inherent nature of the system allows healthcare professionals to rely on its predictions. Additionally, hyperparameter tuning enhances the performance of the model. The robustness of the applied model is evident.