A REVIEW ON KIDNEY STONE DETECTION AND DIAGNOSIS USING CT KUB IMAGING: BRIDGING EMERGING TECHNOLOGIES AND CLINICAL PRACTICE
DOI:
https://doi.org/10.70905/bmcj.06.02.0547Keywords:
Kidney Stones, , Artificial Intelligence, , CT KUB, Deep Learning, Convolution Neural Networks, , Detection, DiagnosisAbstract
Background: Kidney stone disease is a common condition impacting a significant global population, with increasing incidence rates associated with lifestyle changes and dietary factors. Traditional diagnostic methods, including ultrasound and X-rays, have limitations in detecting specific stone types and sizes. CT KUB imaging, however, remains the gold standard for diagnosis.
Objective: This paper aims to evaluate the advancements in artificial intelligence (AI) methods in automating the detection, segmentation, and classification of kidney stones using CT KUB images. Additionally, it investigates the integration of multi-modal data, including patient clinical data, to enhance diagnostic accuracy and treatment personalization.
Material and Methods: The search of the structured literature in PubMed, Scopus, IEEE Xplore, and Google Scholar was conducted in terms of the keywords that included kidney stone, urolithiasis, AI, deep learning, and CT and were published in the period between January 2015 and March 2025. The inclusion criteria included peer-reviewed articles that used AI to detect or classify stones on CT KUB. The total number of records that were screened was 482, 58 full texts evaluated and 32 studies were included according to PRISMA guidelines.
Results: The reported diagnostic performances among AI models showed a range of sensitivities of 88 to 98, specificities of 85 to 97 and AUC of 0.85 to 0.96. Improved segmentation accuracy and diagnostic speed was observed to be constant in Convolution Neural Networks (CNNs), hybrid models and radiomics-based frameworks.
Conclusion: Despite the promising potential of AI in clinical settings, challenges such as data scarcity, model interpretability, and generalization remain. The paper proposes future research directions, particularly in combining explainable AI with federated learning technologies, aimed at seamless real-time clinical integration. Emerging technologies have the potential to revolutionize kidney stone management and urological treatment protocols.
