Kidney Magnetic Resonance Imaging (MRI) has long been considered helpful in clinical practice for predicting kidney function and chronic kidney disease (CKD), since MRI scans may produce more comprehensive images of tissues and organs. This extremely desirable method is, however, beyond human perception. We developed a technique for automatically determining CKD status based on deep learning. Here, T2-W MRIs of the kidney are taken for research. The manuscript proposes ResNet18, inspired by the Convolution Neural Network (CNN), for determining CKD state from MRI images. It seeks to increase performance over CNN, Multi-Layered Perceptron (MLP), and MLP with Discrete Wavelet Features (DWT-MLP) due to relying on skip connections that preclude the vanishing gradient problem. Here, ResNet18 excels all others because of its high accuracy, sensitivity, specificity, precision, geometric mean, \(F_1\) -score, and low false positive rate score after computing the mean of the above performance parameters. With a performance score of 87.6%, the Multi-Criteria Decision Analysis (MCDA) technique TOPSIS shows that ResNet18 outperforms all others. Moreover, due to its cost efficiency and noninvasiveness, our proposed deep learning model may be perfectly adapted for kidney screening.