Automated Kidney Anomaly Detection Using Deep Learning and Explainable AI Techniques
Bobba Siva Sankar Reddy, Nelluru Laxmi Prathyusha, Dhulipudi Venkata Karthik, Kayala Vishnukanth, Venkatramaphanikumar Sistla, Venkata Krishna Kishore Kolli · 2025
Accurate diagnosis of kidney abnormalities such as cysts and tumors is essential for timely and effective treatment. This paper describes an in-depth study on recognizing and identifying kidney abnormalities using CT imaging. Convolutional neural sneŧworks are popular due to their excellent ability to extract complex features from large amounts of medical data and achieve better results than conventional methods. The proposed method combines CNN and previous learning models, including VGG16, EfficientNetB0, and MobileNetV2, all optimized to improve performance and robustness across multiple kidney types. The model shows good specificity in classifying some cases, helping to make more reliable diagnoses. Additionally, descriptive artificial intelligence (XAI) techniques such as LIME and SHAP are used to improve the model by identifying key features that affect the cut-off. This helps increase radiologists’ confidence, improve interpretation, and facilitate clinical decision-making. The results show that this model has the potential to help improve early diagnosis, ultimately improving patient outcomes and diagnostic procedures.