Enhancing Kidney Disease Diagnosis through Advanced Deep Learning Techniques: A Comprehensive Analysis in Modern Medical Imaging

This research work develops an effective approach for diagnosing kidney diseases by applying CNNs. The research focuses on classifying four specific kidney conditions—normal, cyst, tumor, and stone—by training and testing three CNN models: This includes EfficientNetB3, MobileNetV2, and ResNet50. The models were trained on a large set of CT images, and it was noted that the performance parameters were quite high. Among all the EfficientNetB3 models, provided the highest overall accuracy of 90.46%, this indicates greater efficiency in precision and recall in all the categories presented. MobileNetV2 and ResNet50 achieved considerable scores, specifically 89% and 89.50% of accurate predictions, respectively, although there were small discrepancies in their measurements, especially in distinguishing cysts and stones. Resultantly, the study highlights the possibility of CNNs to enhance diagnostic accuracy of kidney disease and consequently support early identification, clinical management, and improved patient safety and prognoses.