Enhancing Renal Stone Detection Through Artificial Intelligence Models

Enhancing Renal Stone Detection Through Artificial Intelligence Models | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Enhancing Renal Stone Detection Through Artificial Intelligence Models Xue Yang, Jian Kang, Jirui Niu, Li Ma, Yin Zhang, Zipu Dong This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9154303/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 20 You are reading this latest preprint version Abstract Purpose Traditional ultrasonic detection for kidney stones is prone to variability due to the experience and skill of the operating physician, leading to inconsistent diagnostic outcomes. Factors such as stone size, location, and surrounding tissues can further compromise accuracy. Artificial intelligence (AI), with its advanced data processing and deep learning capabilities, offers a solution by automating the analysis of medical images to improve diagnostic accuracy and efficiency. Methods This cross-sectional study was conducted at our hospital between 01/06/2022 and 03/07/2024., using a datasets of 791 ultrasound images from patients with confirmed kidney stones and 117 images from non-stone participants. The High-Resolution Network (HRNet) was employed for model training to detect kidney stones in ultrasound images. Detailed image annotations were created using the labelme tool and converted to COCO format for compatibility with AI algorithms. Results The HRNet model achieved an accuracy of 86.4% in detecting kidney stones in the validation datasets, with a sensitivity of 94.3% and specificity of 73.2%. The model effectively identified both kidney stones and normal conditions. Conclusions The study demonstrates that AI, specifically HRNet, can significantly enhance the accuracy and efficiency of kidney stone detection in ultrasound imaging. This approach reduces the diagnostic burden on physicians and improves patient outcomes. However, challenges remain, including the need for large-scale annotated datasets and rigorous validation of AI models to ensure reliability in clinical settings. Ultrasound renal stone artificial intelligence Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Renal stone (RS) is one of the most common urological diseases [ 1 , 2 ] . The incidence of RS has increased dramatically in the last few decades [ 3 ] . Ultrasound detection is one of the most commonly employed methods for diagnosing RS [ 4 , 5 ] . Its non-invasive nature and real-time imaging capabilities make it a preferred choice in clinical settings [ 6 , 7 ] . However, this technique is not without its limitations. The accuracy of ultrasound results is often heavily influenced by the experience and skill level of the operating physician [ 1 – 3 , 8 ] . Variations in interpreting ultrasound images among different doctors can lead to inconsistent diagnostic outcomes, which may significantly impact patient care [ 9 ] . Moreover, the precision of ultrasound detection can be compromised by factors such as the size, location, and surrounding tissues of the stone [ 10 , 11 ] . These factors can obscure the stone or make it difficult to differentiate from other structures, thereby affecting the reliability of the diagnosis [ 12 ] . Additionally, ultrasound detection demands prolonged concentration from doctors, which can easily result in fatigue and misjudgment, further complicating the diagnostic process [ 4 , 6 , 13 ] . With the rapid advancement of technology, artificial intelligence (AI) has emerged as a novel and powerful tool in medical diagnostics [ 14 – 16 ] . AI, particularly through its robust data processing capabilities and deep learning algorithms, offers a promising solution to many of the challenges faced in traditional ultrasound detection of renal stones [ 17 , 18 ] . By training and refining algorithms, AI models can autonomously analyze medical images, pinpoint lesions, and enhance both the accuracy and efficiency of diagnoses [ 8 , 16 , 17 , 19 ] . This not only accelerates the diagnostic process for doctors but also alleviates their workload, ultimately delivering superior medical services to patients. The significance of AI in medical research is becoming increasingly pronounced, especially in the realm of kidney stone detection. In recent years, a growing number of researchers have begun exploring the application of AI technology for the detection of renal stones. Initial studies have shown promising results, demonstrating that AI models can successfully implement automated analysis of ultrasound images and identify renal stones with high accuracy. These models are adept at extracting relevant feature information from images and classifying and identifying features through sophisticated algorithms, thereby facilitating the automatic detection of renal stones. The ability of AI to process large datasets and identify patterns that may be imperceptible to the human eye makes it a valuable tool in improving diagnostic outcomes. However, the integration of AI into the domain of automatic renal stone detection is not without its challenges [ 20 , 21 ] . One of the primary issues is the quality and quantity of data required for training and refining these models. Large-scale, annotated datasets are essential to facilitate effective learning and validation of AI algorithms. The development of such datasets requires significant resources and collaboration between medical professionals and data scientists. Additionally, the reliability and stability of AI models must undergo rigorous testing and assessment to ensure their efficacy in real-world applications. Ensuring that AI models can consistently and accurately detect renal stones across a wide range of patient demographics and clinical scenarios is crucial for their widespread adoption. To address these challenges and capitalize on the potential of AI, we have incorporated AI technology into the application research of ultrasonic detection for renal stones. Our study aims to develop and validate an AI model that can autonomously analyze ultrasound images to detect renal stones with high accuracy. By leveraging advanced data processing techniques and state-of-the-art AI algorithms, we hope to overcome the limitations of traditional ultrasound detection and provide a more reliable and efficient diagnostic tool for clinicians. This research not only seeks to improve the accuracy of renal stone detection but also aims to reduce the burden on medical professionals, ultimately leading to better patient outcomes. Methods Study design and population A cross-sectional study was conducted at Heilongjiang Provincial Hospital between 01/06/2022 and 03/07/2024. The datasets consisted of 791 randomly selected ultrasound images from patients with confirmed renal stones and 117 images from participants without stones. The study population included adult patients who had been diagnosed with at least one renal stone via ultrasound imaging. Ultrasound examinations were performed as part of routine clinical care. The study aimed to evaluate the effectiveness of an AI model in detecting renal stones using ultrasound images. Figures 1 and 2 illustrated the AI model process and the detailed methodology flowchart, respectively. The requirement for obtaining signed informed consent from participants or their guardians was waived due to the retrospective nature of the study and the use of anonymous data. The overall flow chart of the model was shown in Fig. 1 . Data collection Ultrasound image data were meticulously collected from patients with confirmed renal stones and non-stone participants. The data collection process involved capturing high-quality ultrasound images using standardized protocols to ensure consistency and accuracy. The images were securely stored in a digital repository for subsequent processing and sharing. Advanced data processing techniques were employed to ensure the accuracy and efficiency of medical information, including data cleaning and normalization to reduce noise and enhance image quality. Data labeling To enhance the readability and practicality of the data, the professional image annotation tool named labelme, was utilized to conduct detailed annotations on the critical regions of the images. This process involved identifying and marking the location and size of renal stones within the images. Annotations were performed by experienced medical professionals to ensure accuracy. This step not only provided intuitive visual references for researchers but also supplied essential data support for subsequent analyses and the training of AI models. Data exploration and pre-processing The annotated data were transformed into the COCO (Common Objects in Context) format to align with the specifications of diverse systems and algorithms. COCO is a widely-adopted format for object detection and image segmentation, known for its standardized data structure. This format facilitated seamless integration of the data into various analysis and processing systems, enhancing the data's versatility and unlocking new avenues for research on the diagnosis and treatment of kidney stones. Data exploration involved analyzing the distribution of stone sizes, locations, and image quality to identify any potential biases or limitations in the datasets. Feature selection Feature selection was a critical step in preparing the data for AI model training. Relevant features that could aid in the detection of renal stones were extracted. These features included stone size, shape, echogenicity, and surrounding tissue characteristics. The selection process involved both manual feature extraction and automated feature selection techniques to identify the most informative features for model training. Machine learning models The High-Resolution Network (HRNet) was selected as the primary machine learning model for this study [ 22 ] . HRNet is designed to maintain high-resolution representations throughout the entire process, making it particularly suitable for tasks requiring detailed image analysis [ 23 ] . HRNet achieves cross-resolution information interaction by parallelly connecting multi-resolution subnetworks and repeatedly exchanging information among them [ 24 ] . This architecture enables HRNet to maintain high-resolution feature representations and perform multi-scale feature fusion, significantly enhancing the efficiency and accuracy of the detection process [ 25 , 26 ] . Stage 1 begins with a single high-resolution branch, processing the input image through convolutional layers to generate a detailed feature map. Stage 2 introduces a second branch with lower resolution, capturing coarser features. These two branches exchange information, allowing the high-resolution branch to benefit from coarser details and vice versa. Stage 3 adds a third branch with even lower resolution, continuing the information exchange among all three branches to ensure a comprehensive multi-scale representation. Stage 4 may include additional lower-resolution branches, maintaining the principle of cross-resolution interaction. This architecture ensures that HRNet captures both fine-grained and coarser features, enhancing its ability to perform detailed image analysis. This design is particularly effective for tasks requiring high accuracy and robustness, such as medical imaging, where maintaining detailed representations is crucial for tasks like detecting renal stones in ultrasound images. Probability estimate stratification In multiple stages of HRNet, multiple lower-resolution image features were gradually combined to improve the proportion and thickness of high-resolution feature maps. This process enhanced the model's ability to capture fine-grained details and improve detection accuracy. The optimal combination of image features was determined through iterative training and validation. Finally, all feature maps were combined proportionally to generate the final high-resolution image, which was used to estimate the probability of the presence of renal stones. The flowchart is shown in Fig. 2 . Statistical Methods For binary classification tasks, accuracy was calculated as the fraction of correct predictions in all predictions: Accuracy= \(\:\frac{TP+TN}{TP+FP+FN+TN}\) Where TP is true positive, TN is true negative, FP is false positive, and FN is false negative. The confusion matrix, which has four cells (TP, FN, FP, TN), was also used to evaluate the performance of the model. Sensitivity and specificity were calculated as follows: Sensitivity= \(\:\frac{TP}{FN+TP}\) Specificity= \(\:\frac{TN}{FP+TN}\) As shown in the confusion matrix, the model accurately identified both renal stones and normal (non-stone) conditions. Results The AI model for renal stone detection was evaluated and achieved an accuracy of 86.4% in the validation datasets, indicating effective overall diagnostic performance. The confusion matrix analysis was conducted and revealed high sensitivity (94.3%) and moderate specificity (73.2%). The visualization of renal stones was presented in Fig. 3 . An ultrasound image of the kidney was displayed in Fig. 3 A, while the application of AI modules in detecting renal stones was illustrated in Fig. 3 B. A significant proportion of renal stones could be accurately diagnosed by the AI modules, as indicated by the results, which demonstrated their potential effectiveness in clinical settings. The performance of the model in identifying renal stones and non-stone cases was evaluated using the confusion matrix of the test cohort, as shown in Fig. 4 . Discussion Significance of AI in Renal Stone Detection The incorporation of artificial intelligence (AI) into medical diagnostics has shown remarkable potential for enhancing the accuracy and efficiency of renal stone detection [ 16 , 17 , 27 , 28 ] . Traditional ultrasonic methods, despite their widespread use, are subject to variability based on the experience and skill of the operating physician [ 6 ] . This variability can lead to inconsistent diagnostic outcomes, especially when influenced by factors such as stone size, location, and surrounding tissues. The application of AI, particularly the High-Resolution Network (HRNet) in this study, addresses these limitations by automating the analysis of ultrasound images and providing consistent, high-accuracy detection of renal stones [ 12 , 13 , 23 , 29 – 33 ] . AI's ability to process large volumes of complex data and identify subtle patterns that may escape human interpretation is crucial for renal stone detection [ 32 , 33 ] . The variability in stone characteristics and surrounding tissues can obscure findings, making it challenging to identify renal stones accurately [ 13 , 18 ] . By leveraging advanced image classification and segmentation techniques, AI can enhance the quality and resolution of medical images, thereby improving the identification and differentiation of renal stones. Moreover, AI-driven systems can significantly reduce human error in medical imaging diagnostics [ 16 – 18 , 33 , 34 ] . Unlike human clinicians, AI algorithms do not experience fatigue or variability in performance, ensuring consistent and reliable results across different cases [ 33 ] . This consistency is essential in clinical settings, where accurate and timely detection of renal stones is crucial for effective treatment and management [ 18 ] . In addition to improving diagnostic accuracy, AI also enhances the efficiency of the diagnostic process [ 15 , 16 , 20 , 25 , 35 ] . By automating the analysis of ultrasound images, AI models can quickly identify potential renal stones, reducing the time required for diagnosis and allowing clinicians to focus on patient care [ 13 , 19 , 27 , 31 , 32 , 36 ] . This not only improves patient outcomes but also optimizes the overall workflow in medical settings. Overall, the application of AI in renal stone detection represents a significant advancement in medical diagnostics. By addressing the limitations of traditional ultrasonic methods and providing consistent, high-accuracy detection, AI has the potential to revolutionize the way renal stones are identified and managed in clinical practice. Model Performance and Clinical Implications The HRNet model achieved an accuracy of 86.4% in detecting renal stones in the validation datasets, with a sensitivity of 94.3% and specificity of 73.2%. These results indicate that the model is highly effective in identifying the presence of renal stones, ensuring that few cases are missed. This high sensitivity is crucial in clinical settings, as it minimizes the risk of undiagnosed stones, which can lead to complications if left untreated. However, the moderate specificity suggests that the model may generate some false positives, which could lead to unnecessary further investigations or treatments. This highlights the need for further optimization of the model to improve its ability to distinguish between stones and non-stone cases accurately. The use of HRNet in this study demonstrates the potential of AI to significantly enhance the diagnostic process. By maintaining high-resolution representations throughout the network, HRNet is able to capture fine-grained details in images, making it particularly well-suited for tasks requiring detailed image analysis. This capability is essential for the accurate detection of renal stones, which can vary in size and shape and may be obscured by surrounding tissues. Challenges and Future Directions Despite the promising results, several challenges remain in the application of AI for renal stone detection [ 13 , 31 ] . One of the primary issues is the quality and quantity of data required for training and refining AI models [ 37 ] . Large-scale, annotated datasets are essential for effective learning and validation of AI algorithms [ 10 , 38 , 39 ] . The development of such datasets requires significant resources and collaboration between medical professionals and data scientists. Ensuring that AI models can consistently and accurately detect renal stones across a wide range of patient demographics and clinical scenarios is crucial for their widespread adoption. Another challenge is the need for rigorous testing and validation of AI models to ensure their reliability and stability in real-world applications. The model's performance must be validated on larger and more diverse datasets to ensure its generalizability and clinical applicability. Additionally, the integration of AI into clinical workflows requires careful consideration of ethical and practical implications, including patient privacy and data security. Limitations of the Study While the results of this study are encouraging, there are several limitations that must be acknowledged. First, the datasets used in this study was relatively small and derived from a single institution. This limits the generalizability of the findings to broader patient populations. Future studies should include larger, multi-center datasets to ensure that the model performs consistently across diverse populations and clinical settings. Second, the specificity of the model was moderate, indicating a potential for false positives. This could lead to unnecessary further investigations or treatments, which can be both costly and anxiety-inducing for patients. Further refinement of the feature selection process and the development of more sophisticated algorithms are needed to improve specificity without compromising sensitivity. Third, the study relied on ultrasound images, which are inherently subject to variability based on the experience and skill level of the operating physician. While AI can mitigate some of these issues, the quality of the input data remains a critical factor. Future work should explore the integration of AI with other imaging modalities, such as computed tomography (CT), to further enhance diagnostic accuracy. Finally, the study did not address the practical and ethical considerations of integrating AI into clinical workflows. Ensuring patient privacy and data security, as well as addressing potential biases in AI models, are essential for gaining clinical acceptance and ensuring equitable access to care. Conclusions This study demonstrates the potential of AI, specifically HRNet, in enhancing the detection of renal stones using ultrasound images. The model achieved high sensitivity and acceptable accuracy, indicating its effectiveness in identifying renal stones. However, further optimization is needed to improve specificity and reduce false positives. The integration of AI into renal stone detection offers a promising solution to the limitations of traditional ultrasonic methods, with the potential to improve diagnostic outcomes and reduce the burden on medical professionals. Future work will focus on refining the model and validating its performance on larger datasets to enhance its clinical applicability and ensure its widespread adoption in clinical practice. By addressing the remaining challenges and limitations, we can pave the way for more accurate, efficient, and reliable detection of renal stones, ultimately leading to better patient care and outcomes. Declarations Ethics approval and consent to participate The Ethics Committee of the Heilongjiang Provincial Hospital approved this investigation, which was conducted in accordance with the Declaration of Helsinki. Due to the retrospective nature of the study, the Ethics Committee of Heilongjiang Provincial Hospital waived the need of obtaining informed consent. Funding Declaration This study was supported by the scientific research of the Heilongjiang Provincial Health Commission [grant number 20240404050045] and the Heilongjiang Provincial Natural Science Foundation [grant number LH2024H070]. Authors’ contributions XY, LM, JN and JK drafted the manuscript. JN analyzed the data using AI models. JN and LM drew the pictures. ZD, JK, LM and YZ collected the data. ZD reviewed and revised the manuscript. All the authors have read and approved the final version of the manuscript. All the authors contributed to the manuscript and approved the submitted version. Consent for publication Not applicable. Competing interests There is no competing interests. Availability of data and materials The datasets and Python source code for AI model training are available at https://figshare.com/s/13ca30d3149ed5690e23 . References Yao L, Yang P. Relationship between remnant cholesterol and risk of kidney stones in U.S. Adults: a 2007-2016 NHANES analysis. Ann Med. 2024;56(1):2319749. Ferraro PM, Taylor EN, Curhan GC. 24-Hour Urinary Chemistries and Kidney Stone Risk. Am J Kidney Dis. 2024;84(2):164-169. Tamborino F, Cicchetti R, Mascitti M, Litterio G, Orsini A, Ferretti S, Basconi M, De Palma A, Ferro M, Marchioni M, Schips L. Pathophysiology and Main Molecular Mechanisms of Urinary Stone Formation and Recurrence. Int J Mol Sci. 2024;25(5). Xu Z, Li Z, Guo M, Bian H, Niu T, Wang J. Application of Three-dimensional Visualization Fused with Ultrasound for Percutaneous Renal Puncture. Sci Rep. 2021;11(1):8521. Ciccarese F, Brandi N, Corcioni B, Golfieri R, Gaudiano C. Complicated pyelonephritis associated with chronic renal stone disease. Radiol Med. 2021;126(4):505-516. Fontenelle LF, Sarti TD. Kidney Stones: Treatment and Prevention. Am Fam Physician. 2019;99(8):490-496. Armas-Phan M, Tzou DT, Bayne DB, Wiener SV, Stoller ML, Chi T. Ultrasound guidance can be used safely for renal tract dilatation during percutaneous nephrolithotomy. BJU Int. 2020;125(2):284-291. Raffone A, Raimondo D, Neola D, Travaglino A, Raspollini A, Giorgi M, Santoro A, De Meis L, Zannoni GF, Seracchioli R, Casadio P, Guida M. Diagnostic Accuracy of Ultrasound in the Diagnosis of Uterine Leiomyomas and Sarcomas. J Minim Invasive Gynecol. 2024;31(1):28-36.e21. Raja AS, Pourjabbar S, Ip IK, Baugh CW, Sodickson AD, O'Leary M, Khorasani R. Impact of a Health Information Technology-Enabled Appropriate Use Criterion on Utilization of Emergency Department CT for Renal Colic. AJR Am J Roentgenol. 2019;212(1):142-145. Dai JC, Bailey MR, Sorensen MD, Harper JD. Innovations in Ultrasound Technology in the Management of Kidney Stones. Urol Clin North Am. 2019;46(2):273-285. De Perrot T, Hofmeister J, Burgermeister S, Martin SP, Feutry G, Klein J, Montet X. Differentiating kidney stones from phleboliths in unenhanced low-dose computed tomography using radiomics and machine learning. Eur Radiol. 2019;29(9):4776-4782. Längkvist M, Jendeberg J, Thunberg P, Loutfi A, Lidén M. Computer aided detection of ureteral stones in thin slice computed tomography volumes using Convolutional Neural Networks. Comput Biol Med. 2018;97:153-160. Wood BG, Urban MW. Detecting Kidney Stones Using Twinkling Artifacts: Survey of Kidney Stones with Varying Composition and Size. Ultrasound Med Biol. 2020;46(1):156-166. Ahmed F, Abbas S, Athar A, Shahzad T, Khan WA, Alharbi M, Khan MA, Ahmed A. Identification of kidney stones in KUB X-ray images using VGG16 empowered with explainable artificial intelligence. Sci Rep. 2024;14(1):6173. Babajide R, Lembrikova K, Ziemba J, Ding J, Li Y, Fermin AS, Fan Y, Tasian GE. Automated Machine Learning Segmentation and Measurement of Urinary Stones on CT Scan. Urology. 2022;169:41-46. Checcucci E, Autorino R, Cacciamani GE, Amparore D, De Cillis S, Piana A, Piazzolla P, Vezzetti E, Fiori C, Veneziano D, Tewari A, Dasgupta P, Hung A, Gill I, Porpiglia F. Artificial intelligence and neural networks in urology: current clinical applications. Minerva Urol Nefrol. 2020;72(1):49-57. Chen AB, Haque T, Roberts S, Rambhatla S, Cacciamani G, Dasgupta P, Hung AJ. Artificial Intelligence Applications in Urology: Reporting Standards to Achieve Fluency for Urologists. Urol Clin North Am. 2022;49(1):65-117. Shabaniyan T, Parsaei H, Aminsharifi A, Movahedi MM, Jahromi AT, Pouyesh S, Parvin H. An artificial intelligence-based clinical decision support system for large kidney stone treatment. Australas Phys Eng Sci Med. 2019;42(3):771-779. Yildirim K, Bozdag PG, Talo M, Yildirim O, Karabatak M, Acharya UR. Deep learning model for automated kidney stone detection using coronal CT images. Comput Biol Med. 2021;135:104569. Cui Y, Sun Z, Ma S, Liu W, Wang X, Zhang X, Wang X. Automatic Detection and Scoring of Kidney Stones on Noncontrast CT Images Using S.T.O.N.E. Nephrolithometry: Combined Deep Learning and Thresholding Methods. Mol Imaging Biol. 2021;23(3):436-445. Ji Y, Hwang G, Lee SJ, Lee K, Yoon H. A deep learning model for automated kidney calculi detection on non-contrast computed tomography scans in dogs. Front Vet Sci. 2023;10:1236579. Zhao S, Bai Z, Meng L, Han G, Duan E. Pose Estimation and Behavior Classification of Jinling White Duck Based on Improved HRNet. Animals (Basel). 2023;13(18). Wu Y, Chen X, Dong F, He L, Cheng G, Zheng Y, Ma C, Yao H, Zhou S. Performance evaluation of a deep learning-based cascaded HRNet model for automatic measurement of X-ray imaging parameters of lumbar sagittal curvature. Eur Spine J. 2023. Wang G, Zhang S, Li J, Zhao K, Ding Q, Tian D, Li R, Zou F, Yu Q. CB-HRNet: A Class-Balanced High-Resolution Network for the evaluation of endoscopic activity in patients with ulcerative colitis. Clin Transl Sci. 2023;16(8):1421-1430. Ahmed S, Hossain T, Hoque OB, Sarker S, Rahman S, Shah FM. Automated COVID-19 Detection from Chest X-Ray Images: A High-Resolution Network (HRNet) Approach. SN Comput Sci. 2021;2(4):294. Bao W, Niu T, Wang N, Yang X. Pose estimation and motion analysis of ski jumpers based on ECA-HRNet. Sci Rep. 2023;13(1):6132. Caglayan A, Horsanali MO, Kocadurdu K, Ismailoglu E, Guneyli S. Deep learning model-assisted detection of kidney stones on computed tomography. Int Braz J Urol. 2022;48(5):830-839. Cellina M, Cè M, Rossini N, Cacioppa LM, Ascenti V, Carrafiello G, Floridi C. Computed Tomography Urography: State of the Art and Beyond. Tomography. 2023;9(3):909-930. Wang J, Sun K, Cheng T, Jiang B, Deng C, Zhao Y, Liu D, Mu Y, Tan M, Wang X, Liu W, Xiao B. Deep High-Resolution Representation Learning for Visual Recognition. IEEE Trans Pattern Anal Mach Intell. 2021;43(10):3349-3364. Sabuncu Ö, Bilgehan B, Kneebone E, Mirzaei O. Effective deep learning classification for kidney stone using axial computed tomography (CT) images. Biomed Tech (Berl). 2023;68(5):481-491. Thapaliya S, Brady SL, Somasundaram E, Anton CG, Coley BD, Towbin AJ, Zhang B, Dillman JR, Trout AT. Detection of urinary tract calculi on CT images reconstructed with deep learning algorithms. Abdom Radiol (NY). 2022;47(1):265-271. Xing Z, Zhu Z, Jiang Z, Zhao J, Chen Q, Xing W, Pan L, Zeng Y, Liu A, Ding J. Automatic Urinary Stone Detection System for Abdominal Non-Enhanced CT Images Reduces the Burden on Radiologists. J Imaging Inform Med. 2024;37(2):444-454. Yang B, Veneziano D, Somani BK. Artificial intelligence in the diagnosis, treatment and prevention of urinary stones. Curr Opin Urol. 2020;30(6):782-787. Dai JC, Johnson BA. Artificial intelligence in endourology: emerging technology for individualized care. Curr Opin Urol. 2022;32(4):379-392. Kianian R, Carter M, Finkelshtein I, Eleswarapu SV, Kachroo N. Application of Artificial Intelligence to Patient-Targeted Health Information on Kidney Stone Disease. J Ren Nutr. 2024;34(2):170-176. Islam MN, Hasan M, Hossain MK, Alam MGR, Uddin MZ, Soylu A. Vision transformer and explainable transfer learning models for auto detection of kidney cyst, stone and tumor from CT-radiography. Sci Rep. 2022;12(1):11440. Elton DC, Turkbey EB, Pickhardt PJ, Summers RM. A deep learning system for automated kidney stone detection and volumetric segmentation on noncontrast CT scans. Med Phys. 2022;49(4):2545-2554. Li AY, Elliot N. Natural language processing to identify ureteric stones in radiology reports. J Med Imaging Radiat Oncol. 2019;63(3):307-310. Mukherjee P, Lee S, Elton DC, Nakada SY, Pickhardt PJ, Summers RM. Fully Automated Longitudinal Assessment of Renal Stone Burden on Serial CT Imaging Using Deep Learning. J Endourol. 2023;37(8):948-955. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 02 May, 2026 Reviews received at journal 29 Apr, 2026 Reviews received at journal 28 Apr, 2026 Reviews received at journal 26 Apr, 2026 Reviews received at journal 25 Apr, 2026 Reviewers agreed at journal 24 Apr, 2026 Reviews received at journal 22 Apr, 2026 Reviewers agreed at journal 22 Apr, 2026 Reviews received at journal 21 Apr, 2026 Reviewers agreed at journal 19 Apr, 2026 Reviewers agreed at journal 18 Apr, 2026 Reviewers agreed at journal 17 Apr, 2026 Reviewers agreed at journal 17 Apr, 2026 Reviewers agreed at journal 15 Apr, 2026 Reviewers agreed at journal 15 Apr, 2026 Reviewers invited by journal 15 Apr, 2026 Editor invited by journal 24 Mar, 2026 Editor assigned by journal 23 Mar, 2026 Submission checks completed at journal 23 Mar, 2026 First submitted to journal 17 Mar, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9154303","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":627359056,"identity":"8efcfe5e-ad78-49e6-8977-869a2dc3bfc6","order_by":0,"name":"Xue Yang","email":"","orcid":"","institution":"Fourth Affiliated Hospital of Harbin Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xue","middleName":"","lastName":"Yang","suffix":""},{"id":627359057,"identity":"84bfcb0c-883d-4b37-b1b8-45fc0e2344be","order_by":1,"name":"Jian Kang","email":"","orcid":"","institution":"Heilongjiang Provincial Hospital","correspondingAuthor":false,"prefix":"","firstName":"Jian","middleName":"","lastName":"Kang","suffix":""},{"id":627359059,"identity":"79c18ab4-9cc6-4b63-bb77-a82e2c020407","order_by":2,"name":"Jirui Niu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0klEQVRIiWNgGAWjYDACZgYGgwQGBh5+CQY2iMgBYrVIziBaCwwY3CBWizk7j0HBwz2HZYxv9z57dLONQY7vRgLj5wI8WiybeQwMEp4d5jG7c9zcOLeNwVjyRgKz9Ax87jkM0nIAqOVGGps0UEvihhsJbMw8xGgxngHRUk+8FgMJiJYEA8Ja2AqAWtJ5JG6ksRvnnJMwnHnmYbM0Xi3nD28z/HHA2p4f6LDHOWU28nzHkw9+xqcFCNgMkDgSQMzYgF8DMP4fEFIxCkbBKBgFIxwAAH/ERiVULKL7AAAAAElFTkSuQmCC","orcid":"","institution":"Fourth Affiliated Hospital of Harbin Medical University","correspondingAuthor":true,"prefix":"","firstName":"Jirui","middleName":"","lastName":"Niu","suffix":""},{"id":627359060,"identity":"a8cede0c-5840-4f79-a22f-c6e4892dae42","order_by":3,"name":"Li Ma","email":"","orcid":"","institution":"Heilongjiang Nursing College","correspondingAuthor":false,"prefix":"","firstName":"Li","middleName":"","lastName":"Ma","suffix":""},{"id":627359063,"identity":"559f3358-0b67-408f-ab1b-71dac0df96df","order_by":4,"name":"Yin Zhang","email":"","orcid":"","institution":"Heilongjiang Provincial Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yin","middleName":"","lastName":"Zhang","suffix":""},{"id":627359065,"identity":"db54814c-b86a-4baa-982f-84a4a3cda013","order_by":5,"name":"Zipu Dong","email":"","orcid":"","institution":"Heilongjiang Provincial Hospital","correspondingAuthor":false,"prefix":"","firstName":"Zipu","middleName":"","lastName":"Dong","suffix":""}],"badges":[],"createdAt":"2026-03-18 04:08:21","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9154303/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9154303/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107619015,"identity":"459c1007-23fd-4725-af9e-a6b74d5c4653","added_by":"auto","created_at":"2026-04-23 09:27:09","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":123294,"visible":true,"origin":"","legend":"\u003cp\u003eTitle: The overall flow chart of the model.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLegend\u003c/strong\u003e: It presents a detailed flowchart, which initially included 1180 participants, among whom 1063 were diagnosed with kidney stones, while the remaining 117 were healthy participants. To ensure data accuracy, we excluded 50 samples with poor ultrasound quality. Subsequently, we carefully divided the remaining datasets into two separate parts: the first datasets contained 791 samples for model training and validation, while the second datasets comprised 222 kidney stone patients and 117 healthy control participants, specifically for independent validation testing of the AI model. Using the first datasets, we conducted model training and successfully developed an efficient AI model. Then, leveraging the second datasets, we conducted independent validation of the model to ensure its accuracy and reliability.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-9154303/v1/98e148db5c023b10544d7798.png"},{"id":107618966,"identity":"5d3b21cb-7077-475e-94dc-f4865b8bc1d5","added_by":"auto","created_at":"2026-04-23 09:26:57","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":294126,"visible":true,"origin":"","legend":"\u003cp\u003eTitle: The overall pipeline of the AI model.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLegend\u003c/strong\u003e: The diagram demonstrates the process of the AI model from data collection to model establishment. The data is converted into the COCO datasets, and then the HRNET module is applied for training to build the AI model. This model is then used to make judgments with data from additional datasets.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-9154303/v1/50179f1a70dff3661e45fe8a.png"},{"id":107618967,"identity":"7c7dd17b-ad79-4b27-84a2-b734d711293b","added_by":"auto","created_at":"2026-04-23 09:26:57","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":336733,"visible":true,"origin":"","legend":"\u003cp\u003eTitle: Visualization of renal stone.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLegend\u003c/strong\u003e: Fig. 3A is an ultrasound image, and Fig. 3B is a visualization of the application of AI modules. The image shows the percentage of stones that can be diagnosed.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-9154303/v1/cf10a24d7252089302277f90.png"},{"id":107619027,"identity":"6ee083af-fdb4-41a1-bf69-efdff6204466","added_by":"auto","created_at":"2026-04-23 09:27:12","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":40036,"visible":true,"origin":"","legend":"\u003cp\u003eTitle: The confusion Matrix of the test cohort.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLegend\u003c/strong\u003e: The figure depicts a confusion matrix, where the horizontal axis represents the predicted values and the vertical axis represents the actual values. The deeper colors indicate a higher concentration of data points. The matrix suggests that there are a significant number of correct predictions for renal stones as well as for non-stone cases. This indicates that the model is performing well in identifying both conditions accurately.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-9154303/v1/df424fcf736d5016d8594f37.png"},{"id":107707263,"identity":"1d884c5e-42d1-446e-a010-93a94a0add89","added_by":"auto","created_at":"2026-04-24 09:19:55","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1029848,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9154303/v1/3765bf72-c2c6-4b6d-a2b5-734b9d162cd1.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Enhancing Renal Stone Detection Through Artificial Intelligence Models","fulltext":[{"header":"Introduction","content":"\u003cp\u003eRenal stone (RS) is one of the most common urological diseases\u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]\u003c/sup\u003e. The incidence of RS has increased dramatically in the last few decades\u003csup\u003e[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/sup\u003e. Ultrasound detection is one of the most commonly employed methods for diagnosing RS\u003csup\u003e[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]\u003c/sup\u003e. Its non-invasive nature and real-time imaging capabilities make it a preferred choice in clinical settings\u003csup\u003e[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e. However, this technique is not without its limitations. The accuracy of ultrasound results is often heavily influenced by the experience and skill level of the operating physician\u003csup\u003e[\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]\u003c/sup\u003e. Variations in interpreting ultrasound images among different doctors can lead to inconsistent diagnostic outcomes, which may significantly impact patient care\u003csup\u003e[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/sup\u003e. Moreover, the precision of ultrasound detection can be compromised by factors such as the size, location, and surrounding tissues of the stone\u003csup\u003e[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]\u003c/sup\u003e. These factors can obscure the stone or make it difficult to differentiate from other structures, thereby affecting the reliability of the diagnosis\u003csup\u003e[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]\u003c/sup\u003e. Additionally, ultrasound detection demands prolonged concentration from doctors, which can easily result in fatigue and misjudgment, further complicating the diagnostic process\u003csup\u003e[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eWith the rapid advancement of technology, artificial intelligence (AI) has emerged as a novel and powerful tool in medical diagnostics\u003csup\u003e[\u003cspan additionalcitationids=\"CR15\" citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/sup\u003e. AI, particularly through its robust data processing capabilities and deep learning algorithms, offers a promising solution to many of the challenges faced in traditional ultrasound detection of renal stones\u003csup\u003e[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]\u003c/sup\u003e. By training and refining algorithms, AI models can autonomously analyze medical images, pinpoint lesions, and enhance both the accuracy and efficiency of diagnoses\u003csup\u003e[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/sup\u003e. This not only accelerates the diagnostic process for doctors but also alleviates their workload, ultimately delivering superior medical services to patients. The significance of AI in medical research is becoming increasingly pronounced, especially in the realm of kidney stone detection.\u003c/p\u003e \u003cp\u003eIn recent years, a growing number of researchers have begun exploring the application of AI technology for the detection of renal stones. Initial studies have shown promising results, demonstrating that AI models can successfully implement automated analysis of ultrasound images and identify renal stones with high accuracy. These models are adept at extracting relevant feature information from images and classifying and identifying features through sophisticated algorithms, thereby facilitating the automatic detection of renal stones. The ability of AI to process large datasets and identify patterns that may be imperceptible to the human eye makes it a valuable tool in improving diagnostic outcomes.\u003c/p\u003e \u003cp\u003eHowever, the integration of AI into the domain of automatic renal stone detection is not without its challenges\u003csup\u003e[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003e. One of the primary issues is the quality and quantity of data required for training and refining these models. Large-scale, annotated datasets are essential to facilitate effective learning and validation of AI algorithms. The development of such datasets requires significant resources and collaboration between medical professionals and data scientists. Additionally, the reliability and stability of AI models must undergo rigorous testing and assessment to ensure their efficacy in real-world applications. Ensuring that AI models can consistently and accurately detect renal stones across a wide range of patient demographics and clinical scenarios is crucial for their widespread adoption.\u003c/p\u003e \u003cp\u003eTo address these challenges and capitalize on the potential of AI, we have incorporated AI technology into the application research of ultrasonic detection for renal stones. Our study aims to develop and validate an AI model that can autonomously analyze ultrasound images to detect renal stones with high accuracy. By leveraging advanced data processing techniques and state-of-the-art AI algorithms, we hope to overcome the limitations of traditional ultrasound detection and provide a more reliable and efficient diagnostic tool for clinicians. This research not only seeks to improve the accuracy of renal stone detection but also aims to reduce the burden on medical professionals, ultimately leading to better patient outcomes.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy design and population\u003c/h2\u003e \u003cp\u003eA cross-sectional study was conducted at Heilongjiang Provincial Hospital between 01/06/2022 and 03/07/2024. The datasets consisted of 791 randomly selected ultrasound images from patients with confirmed renal stones and 117 images from participants without stones. The study population included adult patients who had been diagnosed with at least one renal stone via ultrasound imaging. Ultrasound examinations were performed as part of routine clinical care. The study aimed to evaluate the effectiveness of an AI model in detecting renal stones using ultrasound images. Figures\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e and \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e illustrated the AI model process and the detailed methodology flowchart, respectively. The requirement for obtaining signed informed consent from participants or their guardians was waived due to the retrospective nature of the study and the use of anonymous data. The overall flow chart of the model was shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eData collection\u003c/h3\u003e\n\u003cp\u003eUltrasound image data were meticulously collected from patients with confirmed renal stones and non-stone participants. The data collection process involved capturing high-quality ultrasound images using standardized protocols to ensure consistency and accuracy. The images were securely stored in a digital repository for subsequent processing and sharing. Advanced data processing techniques were employed to ensure the accuracy and efficiency of medical information, including data cleaning and normalization to reduce noise and enhance image quality.\u003c/p\u003e\n\u003ch3\u003eData labeling\u003c/h3\u003e\n\u003cp\u003eTo enhance the readability and practicality of the data, the professional image annotation tool named labelme, was utilized to conduct detailed annotations on the critical regions of the images. This process involved identifying and marking the location and size of renal stones within the images. Annotations were performed by experienced medical professionals to ensure accuracy. This step not only provided intuitive visual references for researchers but also supplied essential data support for subsequent analyses and the training of AI models.\u003c/p\u003e\n\u003ch3\u003eData exploration and pre-processing\u003c/h3\u003e\n\u003cp\u003eThe annotated data were transformed into the COCO (Common Objects in Context) format to align with the specifications of diverse systems and algorithms. COCO is a widely-adopted format for object detection and image segmentation, known for its standardized data structure. This format facilitated seamless integration of the data into various analysis and processing systems, enhancing the data's versatility and unlocking new avenues for research on the diagnosis and treatment of kidney stones. Data exploration involved analyzing the distribution of stone sizes, locations, and image quality to identify any potential biases or limitations in the datasets.\u003c/p\u003e\n\u003ch3\u003eFeature selection\u003c/h3\u003e\n\u003cp\u003eFeature selection was a critical step in preparing the data for AI model training. Relevant features that could aid in the detection of renal stones were extracted. These features included stone size, shape, echogenicity, and surrounding tissue characteristics. The selection process involved both manual feature extraction and automated feature selection techniques to identify the most informative features for model training.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eMachine learning models\u003c/h2\u003e \u003cp\u003eThe High-Resolution Network (HRNet) was selected as the primary machine learning model for this study\u003csup\u003e[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/sup\u003e. HRNet is designed to maintain high-resolution representations throughout the entire process, making it particularly suitable for tasks requiring detailed image analysis\u003csup\u003e[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]\u003c/sup\u003e. HRNet achieves cross-resolution information interaction by parallelly connecting multi-resolution subnetworks and repeatedly exchanging information among them\u003csup\u003e[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]\u003c/sup\u003e. This architecture enables HRNet to maintain high-resolution feature representations and perform multi-scale feature fusion, significantly enhancing the efficiency and accuracy of the detection process\u003csup\u003e[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eStage 1 begins with a single high-resolution branch, processing the input image through convolutional layers to generate a detailed feature map. Stage 2 introduces a second branch with lower resolution, capturing coarser features. These two branches exchange information, allowing the high-resolution branch to benefit from coarser details and vice versa. Stage 3 adds a third branch with even lower resolution, continuing the information exchange among all three branches to ensure a comprehensive multi-scale representation. Stage 4 may include additional lower-resolution branches, maintaining the principle of cross-resolution interaction. This architecture ensures that HRNet captures both fine-grained and coarser features, enhancing its ability to perform detailed image analysis. This design is particularly effective for tasks requiring high accuracy and robustness, such as medical imaging, where maintaining detailed representations is crucial for tasks like detecting renal stones in ultrasound images.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eProbability estimate stratification\u003c/h3\u003e\n\u003cp\u003eIn multiple stages of HRNet, multiple lower-resolution image features were gradually combined to improve the proportion and thickness of high-resolution feature maps. This process enhanced the model's ability to capture fine-grained details and improve detection accuracy. The optimal combination of image features was determined through iterative training and validation. Finally, all feature maps were combined proportionally to generate the final high-resolution image, which was used to estimate the probability of the presence of renal stones. The flowchart is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\n\u003ch3\u003eStatistical Methods\u003c/h3\u003e\n\u003cp\u003eFor binary classification tasks, accuracy was calculated as the fraction of correct predictions in all predictions:\u003c/p\u003e \u003cp\u003eAccuracy=\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{TP+TN}{TP+FP+FN+TN}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003cp\u003eWhere TP is true positive, TN is true negative, FP is false positive, and FN is false negative. The confusion matrix, which has four cells (TP, FN, FP, TN), was also used to evaluate the performance of the model. Sensitivity and specificity were calculated as follows:\u003c/p\u003e \u003cp\u003eSensitivity=\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{TP}{FN+TP}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003cp\u003eSpecificity=\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{TN}{FP+TN}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003cp\u003eAs shown in the confusion matrix, the model accurately identified both renal stones and normal (non-stone) conditions.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eThe AI model for renal stone detection was evaluated and achieved an accuracy of 86.4% in the validation datasets, indicating effective overall diagnostic performance. The confusion matrix analysis was conducted and revealed high sensitivity (94.3%) and moderate specificity (73.2%).\u003c/p\u003e \u003cp\u003eThe visualization of renal stones was presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. An ultrasound image of the kidney was displayed in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA, while the application of AI modules in detecting renal stones was illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB. A significant proportion of renal stones could be accurately diagnosed by the AI modules, as indicated by the results, which demonstrated their potential effectiveness in clinical settings.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe performance of the model in identifying renal stones and non-stone cases was evaluated using the confusion matrix of the test cohort, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eSignificance of AI in Renal Stone Detection\u003c/h2\u003e \u003cp\u003eThe incorporation of artificial intelligence (AI) into medical diagnostics has shown remarkable potential for enhancing the accuracy and efficiency of renal stone detection\u003csup\u003e[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]\u003c/sup\u003e. Traditional ultrasonic methods, despite their widespread use, are subject to variability based on the experience and skill of the operating physician\u003csup\u003e[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/sup\u003e. This variability can lead to inconsistent diagnostic outcomes, especially when influenced by factors such as stone size, location, and surrounding tissues. The application of AI, particularly the High-Resolution Network (HRNet) in this study, addresses these limitations by automating the analysis of ultrasound images and providing consistent, high-accuracy detection of renal stones\u003csup\u003e[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan additionalcitationids=\"CR30 CR31 CR32\" citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eAI's ability to process large volumes of complex data and identify subtle patterns that may escape human interpretation is crucial for renal stone detection\u003csup\u003e[\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]\u003c/sup\u003e. The variability in stone characteristics and surrounding tissues can obscure findings, making it challenging to identify renal stones accurately\u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]\u003c/sup\u003e. By leveraging advanced image classification and segmentation techniques, AI can enhance the quality and resolution of medical images, thereby improving the identification and differentiation of renal stones.\u003c/p\u003e \u003cp\u003eMoreover, AI-driven systems can significantly reduce human error in medical imaging diagnostics\u003csup\u003e[\u003cspan additionalcitationids=\"CR17\" citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]\u003c/sup\u003e. Unlike human clinicians, AI algorithms do not experience fatigue or variability in performance, ensuring consistent and reliable results across different cases\u003csup\u003e[\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]\u003c/sup\u003e. This consistency is essential in clinical settings, where accurate and timely detection of renal stones is crucial for effective treatment and management\u003csup\u003e[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn addition to improving diagnostic accuracy, AI also enhances the efficiency of the diagnostic process\u003csup\u003e[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]\u003c/sup\u003e. By automating the analysis of ultrasound images, AI models can quickly identify potential renal stones, reducing the time required for diagnosis and allowing clinicians to focus on patient care\u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]\u003c/sup\u003e. This not only improves patient outcomes but also optimizes the overall workflow in medical settings.\u003c/p\u003e \u003cp\u003eOverall, the application of AI in renal stone detection represents a significant advancement in medical diagnostics. By addressing the limitations of traditional ultrasonic methods and providing consistent, high-accuracy detection, AI has the potential to revolutionize the way renal stones are identified and managed in clinical practice.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eModel Performance and Clinical Implications\u003c/h2\u003e \u003cp\u003eThe HRNet model achieved an accuracy of 86.4% in detecting renal stones in the validation datasets, with a sensitivity of 94.3% and specificity of 73.2%. These results indicate that the model is highly effective in identifying the presence of renal stones, ensuring that few cases are missed. This high sensitivity is crucial in clinical settings, as it minimizes the risk of undiagnosed stones, which can lead to complications if left untreated. However, the moderate specificity suggests that the model may generate some false positives, which could lead to unnecessary further investigations or treatments. This highlights the need for further optimization of the model to improve its ability to distinguish between stones and non-stone cases accurately.\u003c/p\u003e \u003cp\u003eThe use of HRNet in this study demonstrates the potential of AI to significantly enhance the diagnostic process. By maintaining high-resolution representations throughout the network, HRNet is able to capture fine-grained details in images, making it particularly well-suited for tasks requiring detailed image analysis. This capability is essential for the accurate detection of renal stones, which can vary in size and shape and may be obscured by surrounding tissues.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eChallenges and Future Directions\u003c/h2\u003e \u003cp\u003eDespite the promising results, several challenges remain in the application of AI for renal stone detection\u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]\u003c/sup\u003e. One of the primary issues is the quality and quantity of data required for training and refining AI models\u003csup\u003e[\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]\u003c/sup\u003e. Large-scale, annotated datasets are essential for effective learning and validation of AI algorithms\u003csup\u003e[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]\u003c/sup\u003e. The development of such datasets requires significant resources and collaboration between medical professionals and data scientists. Ensuring that AI models can consistently and accurately detect renal stones across a wide range of patient demographics and clinical scenarios is crucial for their widespread adoption.\u003c/p\u003e \u003cp\u003eAnother challenge is the need for rigorous testing and validation of AI models to ensure their reliability and stability in real-world applications. The model's performance must be validated on larger and more diverse datasets to ensure its generalizability and clinical applicability. Additionally, the integration of AI into clinical workflows requires careful consideration of ethical and practical implications, including patient privacy and data security.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eLimitations of the Study\u003c/h2\u003e \u003cp\u003eWhile the results of this study are encouraging, there are several limitations that must be acknowledged. First, the datasets used in this study was relatively small and derived from a single institution. This limits the generalizability of the findings to broader patient populations. Future studies should include larger, multi-center datasets to ensure that the model performs consistently across diverse populations and clinical settings.\u003c/p\u003e \u003cp\u003eSecond, the specificity of the model was moderate, indicating a potential for false positives. This could lead to unnecessary further investigations or treatments, which can be both costly and anxiety-inducing for patients. Further refinement of the feature selection process and the development of more sophisticated algorithms are needed to improve specificity without compromising sensitivity.\u003c/p\u003e \u003cp\u003eThird, the study relied on ultrasound images, which are inherently subject to variability based on the experience and skill level of the operating physician. While AI can mitigate some of these issues, the quality of the input data remains a critical factor. Future work should explore the integration of AI with other imaging modalities, such as computed tomography (CT), to further enhance diagnostic accuracy.\u003c/p\u003e \u003cp\u003eFinally, the study did not address the practical and ethical considerations of integrating AI into clinical workflows. Ensuring patient privacy and data security, as well as addressing potential biases in AI models, are essential for gaining clinical acceptance and ensuring equitable access to care.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusions","content":"\u003cp\u003eThis study demonstrates the potential of AI, specifically HRNet, in enhancing the detection of renal stones using ultrasound images. The model achieved high sensitivity and acceptable accuracy, indicating its effectiveness in identifying renal stones. However, further optimization is needed to improve specificity and reduce false positives. The integration of AI into renal stone detection offers a promising solution to the limitations of traditional ultrasonic methods, with the potential to improve diagnostic outcomes and reduce the burden on medical professionals. Future work will focus on refining the model and validating its performance on larger datasets to enhance its clinical applicability and ensure its widespread adoption in clinical practice. By addressing the remaining challenges and limitations, we can pave the way for more accurate, efficient, and reliable detection of renal stones, ultimately leading to better patient care and outcomes.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003e\u003cem\u003eEthics approval and consent to participate\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe Ethics Committee of the Heilongjiang Provincial Hospital approved this investigation, which was conducted in accordance with the Declaration of Helsinki. Due to the retrospective nature of the study, the Ethics Committee of Heilongjiang Provincial Hospital waived the need of obtaining informed consent.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eFunding Declaration\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was supported by the scientific research of the Heilongjiang\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eProvincial Health Commission [grant number 20240404050045] and the Heilongjiang Provincial Natural Science Foundation [grant number LH2024H070].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eAuthors\u0026rsquo; contributions\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eXY, LM, JN and JK drafted the manuscript. JN analyzed the data using AI models. JN and LM drew the pictures. ZD, JK, LM and YZ collected the data. ZD reviewed and revised the manuscript. All the authors have read and approved the final version of the manuscript. All the authors contributed to the manuscript and approved the submitted version.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eConsent for publication\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eCompeting interests\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThere is no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eAvailability of data and materials\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003cbr\u003eThe datasets and Python source code for AI model training are available at \u003cem\u003ehttps://figshare.com/s/13ca30d3149ed5690e23\u003c/em\u003e\u003cem\u003e.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eYao L, Yang P. Relationship between remnant cholesterol and risk of kidney stones in U.S. Adults: a 2007-2016 NHANES analysis. Ann Med. 2024;56(1):2319749.\u003c/li\u003e\n\u003cli\u003eFerraro PM, Taylor EN, Curhan GC. 24-Hour Urinary Chemistries and Kidney Stone Risk. Am J Kidney Dis. 2024;84(2):164-169.\u003c/li\u003e\n\u003cli\u003eTamborino F, Cicchetti R, Mascitti M, Litterio G, Orsini A, Ferretti S, Basconi M, De Palma A, Ferro M, Marchioni M, Schips L. Pathophysiology and Main Molecular Mechanisms of Urinary Stone Formation and Recurrence. Int J Mol Sci. 2024;25(5).\u003c/li\u003e\n\u003cli\u003eXu Z, Li Z, Guo M, Bian H, Niu T, Wang J. Application of Three-dimensional Visualization Fused with Ultrasound for Percutaneous Renal Puncture. Sci Rep. 2021;11(1):8521.\u003c/li\u003e\n\u003cli\u003eCiccarese F, Brandi N, Corcioni B, Golfieri R, Gaudiano C. Complicated pyelonephritis associated with chronic renal stone disease. Radiol Med. 2021;126(4):505-516.\u003c/li\u003e\n\u003cli\u003eFontenelle LF, Sarti TD. Kidney Stones: Treatment and Prevention. Am Fam Physician. 2019;99(8):490-496.\u003c/li\u003e\n\u003cli\u003eArmas-Phan M, Tzou DT, Bayne DB, Wiener SV, Stoller ML, Chi T. Ultrasound guidance can be used safely for renal tract dilatation during percutaneous nephrolithotomy. BJU Int. 2020;125(2):284-291.\u003c/li\u003e\n\u003cli\u003eRaffone A, Raimondo D, Neola D, Travaglino A, Raspollini A, Giorgi M, Santoro A, De Meis L, Zannoni GF, Seracchioli R, Casadio P, Guida M. Diagnostic Accuracy of Ultrasound in the Diagnosis of Uterine Leiomyomas and Sarcomas. J Minim Invasive Gynecol. 2024;31(1):28-36.e21.\u003c/li\u003e\n\u003cli\u003eRaja AS, Pourjabbar S, Ip IK, Baugh CW, Sodickson AD, O\u0026apos;Leary M, Khorasani R. Impact of a Health Information Technology-Enabled Appropriate Use Criterion on Utilization of Emergency Department CT for Renal Colic. AJR Am J Roentgenol. 2019;212(1):142-145.\u003c/li\u003e\n\u003cli\u003eDai JC, Bailey MR, Sorensen MD, Harper JD. Innovations in Ultrasound Technology in the Management of Kidney Stones. Urol Clin North Am. 2019;46(2):273-285.\u003c/li\u003e\n\u003cli\u003eDe Perrot T, Hofmeister J, Burgermeister S, Martin SP, Feutry G, Klein J, Montet X. Differentiating kidney stones from phleboliths in unenhanced low-dose computed tomography using radiomics and machine learning. Eur Radiol. 2019;29(9):4776-4782.\u003c/li\u003e\n\u003cli\u003eL\u0026auml;ngkvist M, Jendeberg J, Thunberg P, Loutfi A, Lid\u0026eacute;n M. Computer aided detection of ureteral stones in thin slice computed tomography volumes using Convolutional Neural Networks. Comput Biol Med. 2018;97:153-160.\u003c/li\u003e\n\u003cli\u003eWood BG, Urban MW. Detecting Kidney Stones Using Twinkling Artifacts: Survey of Kidney Stones with Varying Composition and Size. Ultrasound Med Biol. 2020;46(1):156-166.\u003c/li\u003e\n\u003cli\u003eAhmed F, Abbas S, Athar A, Shahzad T, Khan WA, Alharbi M, Khan MA, Ahmed A. Identification of kidney stones in KUB X-ray images using VGG16 empowered with explainable artificial intelligence. Sci Rep. 2024;14(1):6173.\u003c/li\u003e\n\u003cli\u003eBabajide R, Lembrikova K, Ziemba J, Ding J, Li Y, Fermin AS, Fan Y, Tasian GE. Automated Machine Learning Segmentation and Measurement of Urinary Stones on CT Scan. Urology. 2022;169:41-46.\u003c/li\u003e\n\u003cli\u003eCheccucci E, Autorino R, Cacciamani GE, Amparore D, De Cillis S, Piana A, Piazzolla P, Vezzetti E, Fiori C, Veneziano D, Tewari A, Dasgupta P, Hung A, Gill I, Porpiglia F. Artificial intelligence and neural networks in urology: current clinical applications. Minerva Urol Nefrol. 2020;72(1):49-57.\u003c/li\u003e\n\u003cli\u003eChen AB, Haque T, Roberts S, Rambhatla S, Cacciamani G, Dasgupta P, Hung AJ. Artificial Intelligence Applications in Urology: Reporting Standards to Achieve Fluency for Urologists. Urol Clin North Am. 2022;49(1):65-117.\u003c/li\u003e\n\u003cli\u003eShabaniyan T, Parsaei H, Aminsharifi A, Movahedi MM, Jahromi AT, Pouyesh S, Parvin H. An artificial intelligence-based clinical decision support system for large kidney stone treatment. Australas Phys Eng Sci Med. 2019;42(3):771-779.\u003c/li\u003e\n\u003cli\u003eYildirim K, Bozdag PG, Talo M, Yildirim O, Karabatak M, Acharya UR. Deep learning model for automated kidney stone detection using coronal CT images. Comput Biol Med. 2021;135:104569.\u003c/li\u003e\n\u003cli\u003eCui Y, Sun Z, Ma S, Liu W, Wang X, Zhang X, Wang X. Automatic Detection and Scoring of Kidney Stones on Noncontrast CT Images Using S.T.O.N.E. Nephrolithometry: Combined Deep Learning and Thresholding Methods. Mol Imaging Biol. 2021;23(3):436-445.\u003c/li\u003e\n\u003cli\u003eJi Y, Hwang G, Lee SJ, Lee K, Yoon H. A deep learning model for automated kidney calculi detection on non-contrast computed tomography scans in dogs. Front Vet Sci. 2023;10:1236579.\u003c/li\u003e\n\u003cli\u003eZhao S, Bai Z, Meng L, Han G, Duan E. Pose Estimation and Behavior Classification of Jinling White Duck Based on Improved HRNet. Animals (Basel). 2023;13(18).\u003c/li\u003e\n\u003cli\u003eWu Y, Chen X, Dong F, He L, Cheng G, Zheng Y, Ma C, Yao H, Zhou S. Performance evaluation of a deep learning-based cascaded HRNet model for automatic measurement of X-ray imaging parameters of lumbar sagittal curvature. Eur Spine J. 2023.\u003c/li\u003e\n\u003cli\u003eWang G, Zhang S, Li J, Zhao K, Ding Q, Tian D, Li R, Zou F, Yu Q. CB-HRNet: A Class-Balanced High-Resolution Network for the evaluation of endoscopic activity in patients with ulcerative colitis. Clin Transl Sci. 2023;16(8):1421-1430.\u003c/li\u003e\n\u003cli\u003eAhmed S, Hossain T, Hoque OB, Sarker S, Rahman S, Shah FM. Automated COVID-19 Detection from Chest X-Ray Images: A High-Resolution Network (HRNet) Approach. SN Comput Sci. 2021;2(4):294.\u003c/li\u003e\n\u003cli\u003eBao W, Niu T, Wang N, Yang X. Pose estimation and motion analysis of ski jumpers based on ECA-HRNet. Sci Rep. 2023;13(1):6132.\u003c/li\u003e\n\u003cli\u003eCaglayan A, Horsanali MO, Kocadurdu K, Ismailoglu E, Guneyli S. Deep learning model-assisted detection of kidney stones on computed tomography. Int Braz J Urol. 2022;48(5):830-839.\u003c/li\u003e\n\u003cli\u003eCellina M, C\u0026egrave; M, Rossini N, Cacioppa LM, Ascenti V, Carrafiello G, Floridi C. Computed Tomography Urography: State of the Art and Beyond. Tomography. 2023;9(3):909-930.\u003c/li\u003e\n\u003cli\u003eWang J, Sun K, Cheng T, Jiang B, Deng C, Zhao Y, Liu D, Mu Y, Tan M, Wang X, Liu W, Xiao B. Deep High-Resolution Representation Learning for Visual Recognition. IEEE Trans Pattern Anal Mach Intell. 2021;43(10):3349-3364.\u003c/li\u003e\n\u003cli\u003eSabuncu \u0026Ouml;, Bilgehan B, Kneebone E, Mirzaei O. Effective deep learning classification for kidney stone using axial computed tomography (CT) images. Biomed Tech (Berl). 2023;68(5):481-491.\u003c/li\u003e\n\u003cli\u003eThapaliya S, Brady SL, Somasundaram E, Anton CG, Coley BD, Towbin AJ, Zhang B, Dillman JR, Trout AT. Detection of urinary tract calculi on CT images reconstructed with deep learning algorithms. Abdom Radiol (NY). 2022;47(1):265-271.\u003c/li\u003e\n\u003cli\u003eXing Z, Zhu Z, Jiang Z, Zhao J, Chen Q, Xing W, Pan L, Zeng Y, Liu A, Ding J. Automatic Urinary Stone Detection System for Abdominal Non-Enhanced CT Images Reduces the Burden on Radiologists. J Imaging Inform Med. 2024;37(2):444-454.\u003c/li\u003e\n\u003cli\u003eYang B, Veneziano D, Somani BK. Artificial intelligence in the diagnosis, treatment and prevention of urinary stones. Curr Opin Urol. 2020;30(6):782-787.\u003c/li\u003e\n\u003cli\u003eDai JC, Johnson BA. Artificial intelligence in endourology: emerging technology for individualized care. Curr Opin Urol. 2022;32(4):379-392.\u003c/li\u003e\n\u003cli\u003eKianian R, Carter M, Finkelshtein I, Eleswarapu SV, Kachroo N. Application of Artificial Intelligence to Patient-Targeted Health Information on Kidney Stone Disease. J Ren Nutr. 2024;34(2):170-176.\u003c/li\u003e\n\u003cli\u003eIslam MN, Hasan M, Hossain MK, Alam MGR, Uddin MZ, Soylu A. Vision transformer and explainable transfer learning models for auto detection of kidney cyst, stone and tumor from CT-radiography. Sci Rep. 2022;12(1):11440.\u003c/li\u003e\n\u003cli\u003eElton DC, Turkbey EB, Pickhardt PJ, Summers RM. A deep learning system for automated kidney stone detection and volumetric segmentation on noncontrast CT scans. Med Phys. 2022;49(4):2545-2554.\u003c/li\u003e\n\u003cli\u003eLi AY, Elliot N. Natural language processing to identify ureteric stones in radiology reports. J Med Imaging Radiat Oncol. 2019;63(3):307-310.\u003c/li\u003e\n\u003cli\u003eMukherjee P, Lee S, Elton DC, Nakada SY, Pickhardt PJ, Summers RM. Fully Automated Longitudinal Assessment of Renal Stone Burden on Serial CT Imaging Using Deep Learning. J Endourol. 2023;37(8):948-955. \u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Ultrasound, renal stone, artificial intelligence","lastPublishedDoi":"10.21203/rs.3.rs-9154303/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9154303/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e \u003cb\u003ePurpose\u003c/b\u003e \u003c/p\u003e \u003cp\u003eTraditional ultrasonic detection for kidney stones is prone to variability due to the experience and skill of the operating physician, leading to inconsistent diagnostic outcomes. Factors such as stone size, location, and surrounding tissues can further compromise accuracy. Artificial intelligence (AI), with its advanced data processing and deep learning capabilities, offers a solution by automating the analysis of medical images to improve diagnostic accuracy and efficiency.\u003c/p\u003e \u003cp\u003e \u003cb\u003eMethods\u003c/b\u003e \u003c/p\u003e \u003cp\u003e This cross-sectional study was conducted at our hospital between 01/06/2022 and 03/07/2024., using a datasets of 791 ultrasound images from patients with confirmed kidney stones and 117 images from non-stone participants. The High-Resolution Network (HRNet) was employed for model training to detect kidney stones in ultrasound images. Detailed image annotations were created using the labelme tool and converted to COCO format for compatibility with AI algorithms.\u003c/p\u003e \u003cp\u003e \u003cb\u003eResults\u003c/b\u003e \u003c/p\u003e \u003cp\u003eThe HRNet model achieved an accuracy of 86.4% in detecting kidney stones in the validation datasets, with a sensitivity of 94.3% and specificity of 73.2%. The model effectively identified both kidney stones and normal conditions.\u003c/p\u003e \u003cp\u003e \u003cb\u003eConclusions\u003c/b\u003e \u003c/p\u003e \u003cp\u003eThe study demonstrates that AI, specifically HRNet, can significantly enhance the accuracy and efficiency of kidney stone detection in ultrasound imaging. This approach reduces the diagnostic burden on physicians and improves patient outcomes. However, challenges remain, including the need for large-scale annotated datasets and rigorous validation of AI models to ensure reliability in clinical settings.\u003c/p\u003e","manuscriptTitle":"Enhancing Renal Stone Detection Through Artificial Intelligence Models","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-23 09:26:08","doi":"10.21203/rs.3.rs-9154303/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-05-02T14:27:10+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-29T15:34:36+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-28T15:56:42+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-26T20:19:53+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-25T09:00:51+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"166327541493618762353025949222416282267","date":"2026-04-24T10:10:54+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-22T14:09:14+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"330421496837622952987789595397635040288","date":"2026-04-22T13:56:52+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-22T00:56:17+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"241385034800075898373146120607417775268","date":"2026-04-19T10:17:07+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"264422805205987033181895088115900979508","date":"2026-04-18T18:12:18+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"286299231475613785803287057333706559124","date":"2026-04-17T18:11:19+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"81325832944356892569986913592491819387","date":"2026-04-17T15:13:43+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"323515527093143171744621809336757987029","date":"2026-04-15T11:10:20+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"150756367104384753308491895037900915952","date":"2026-04-15T09:10:42+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-15T08:53:30+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-03-24T08:34:20+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-03-23T13:03:27+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-23T13:02:30+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Imaging","date":"2026-03-18T03:53:30+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"da0a4f91-13e4-4193-8877-fa32e8300272","owner":[],"postedDate":"April 23rd, 2026","published":true,"recentEditorialEvents":[{"type":"editorInvitedReview","content":"","date":"2026-05-02T14:27:10+00:00","index":69,"fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-29T15:34:36+00:00","index":68,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-04-23T09:26:08+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-23 09:26:08","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9154303","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9154303","identity":"rs-9154303","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}