Role of Artificial Intelligence in Detection of Renal Stones in Plain CT KUB Using a Deep Learning U-Net Model

Renal stone disease is a significant clinical problem requiring accurate imaging for timely diagnosis. Computed Tomography of the Kidney, Ureter, and Bladder (CT KUB) is the imaging modality of choice due to its superior sensitivity. The integration of Artificial Intelligence (AI), particularly deep learning models such as the 3D CNN U-Net, has shown promise in automating the detection and analysis of renal calculi. This study evaluates the diagnostic accuracy of a 3D CNN U-Net model in detecting, localizing, and characterizing renal stones on CT KUB. Among 70 patients with suspected nephrolithiasis, AI achieved a sensitivity of 94.8% and specificity of 82.1%. The model demonstrated high concordance with radiologist interpretations and significantly reduced reporting time. AI-based tools could revolutionize the diagnostic workflow in radiology by enhancing accuracy and efficiency.