Hybrid Deep Learning Models for Early Detection of Chronic Kidney Disease: A 1-D CNN and ResNet 152-V2 Approach

Chronic kidney disease (CKD) is a gradual and irreversible health disorder that can potentially lead to permanent kidney damage and impaired renal function if untreated. Early prediction and detection of CKD are crucial to enhance treatment outcomes and possibly prevent or delay disease progression. Deep learning (DL) techniques have proven to be reliable in early medical diagnosis. Traditional diagnostic techniques such as, Urine Protein-to-Creatinine Ratio (UPCR) test, Serum Creatinine (SCr) test methods are invasive, time-consuming, expensive, and often detect the disease at the advanced stages. This study proposes a hybrid DL method, combining 1- Dimensional Convolutional Neural Network (1-D CNN) and ResNet 152–V2 for identifying temporal features and spatial hierarchies for improved analysis. The hybrid approach significantly improves CKD detection by effectively analyzing sequential patterns and uncovering complex relationships in the data. The study analyses a dataset of 400 patient samples, each categorized by 37 distinct features comprising clinical parameters and patient profiles. The proposed framework yields $98.95 \%$ accuracy, $100 \%$ precision, $98.56 \%$ recall and 99.34% F1 score, significantly outperforming existing models while effectively reducing false negatives and false positives. The ResNet152V2 model exhibits excellent performance in classifying CKD, offering a valuable tool for physicians to diagnose CKD patients, ultimately enhancing patient outcomes and care.