Machine Learning-Based Prediction of Stone-Free Rate After Retrograde Intrarenal Surgery for Lower Pole Renal Stones

Machine Learning-Based Prediction of Stone-Free Rate After Retrograde Intrarenal Surgery for Lower Pole Renal Stones | 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 Machine Learning-Based Prediction of Stone-Free Rate After Retrograde Intrarenal Surgery for Lower Pole Renal Stones Hsiang Ying Lee, Sung Yong Cho, Yu-Hung Tung, Jose Carlo Elises, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6517712/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 12 Jul, 2025 Read the published version in World Journal of Urology → Version 1 posted 8 You are reading this latest preprint version Abstract Background: Lower pole renal stones (LPS) present unique challenges for retrograde intrarenal surgery (RIRS) due to unfavorable anatomical features, often resulting in suboptimal stone-free rates (SFR). Recent advancements in machine learning (ML) offer new opportunities to predict surgical outcomes and guide clinical decision-making. This study aimed to develop and validate ML-based models to predict SFR following RIRS for LPS. Materials and Methods: We retrospectively analyzed data from 327 patients with LPS who underwent RIRS at two academic institutions: Kaohsiung Medical University Hospital (KMUH, n = 193) and Seoul National University Hospital (SNUH, n = 134). Demographic, anatomical, and stone-related variables were collected, including stone burden, Hounsfield unit (HU), pelvic stone angle (PSA), and renal infundibular length (RIL). A Light Gradient Boosting Machine (LightGBM) algorithm was developed using KMUH data and externally validated with SNUH data. SHAP (SHapley Additive exPlanations) analysis was performed to interpret feature importance. Results: The LightGBM model achieved the highest predictive performance. External validation using the SNUH dataset yielded an accuracy of 77.1%, AUC of 0.759, and F1-score of 0.853. SHAP analysis revealed that stone burden, HU, PSA, and RIL were the most influential features. Notably, PSA demonstrated strong predictive relevance, supporting its use as an alternative to the traditional infundibulopelvic angle (IPA) in anatomical assessment. Conclusions: ML-based models, particularly LightGBM, offer robust predictive capability for SFR following RIRS in patients with LPS. These tools may enhance preoperative planning and personalized surgical strategies. Future prospective studies are warranted to further validate their clinical utility and expand on feature integration. Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Lower pole stones (LPS) is the most common position of renal stones, and it comprised about 25%~35% of renal stones ( 1 ). LPS were considered as challenging position for retrograde intrarenal surgery (RIRS) due to the unique anatomical characteristics with difficulty in removing fragments. It can limit access for lithotripsy and make it difficult to extract stone fragments, even after comprehensive treatment ( 2 ). Although laser technology, endoscopic instruments, and lithotripsy accessories have seen significant advancements in recent years, the stone-free rate (SFR) for LPS remains lower compared to stones in other locations ( 3 ). The SFR for LPS treated with flexible ureteroscopy (fURS) remains suboptimal. One major factor is the presence of an acute infundibulopelvic angle (IPA) and a long, narrow infundibulum, both of which hinder access to the lower pole calyces ( 4 – 5 ). Although advancements in ureteroscope technology have improved deflection capabilities, the insertion of instruments such as laser fibers or stone retrieval baskets through the working channel can still significantly limit scope maneuverability ( 6 ). To address these challenges and improve SFR, several techniques have been proposed. These include stone relocation from the lower pole to more accessible calyces ( 7 ), the use of smaller diameter laser fibers or baskets to enhance deflection, and the application of dusting modes that facilitate spontaneous clearance of stone fragments ( 8 ). Additionally, the development and adoption of suction-assisted access sheaths have shown promise in fragment evacuation ( 9 ). Despite these innovations, the implementation of targeted strategies specifically designed to optimize the management of LPS remains essential for improving surgical outcomes. From previous literature, preoperative predictive scoring system or parameters were designed to predict the SFR of LPS. Stephanie L. Dresner et al. ( 10 ) demonstrated the role of IPA. The more acute IPA and larger preoperative stone size negatively affect SFR and need for repeat surgery. Yuleng Huang et al. ( 11 ) propose a scoring system comprised of stone characteristics, collecting system anatomy to predict the SFR of 1–2 cm LPS. Due to the difficulty of position and better decision-making before treatment plan, we conducted a machine-learning (ML) prediction model including different aspects of characteristics for better prediction. Methods and materials We conducted a retrospective analysis of 327 patients with lower pole renal stones who underwent RIRS. Among them, 193 patients were treated at Kaohsiung Medical University Hospital (KMUH) between April 2018 and October 2023, and 134 patients were treated at Seoul National University Hospital (SNUH) from January 2022 to January 2024. Patients with rare or atypical conditions were excluded, as their limited numbers could have introduced bias and affected the generalizability of the results. Inclusion criteria were as follows: ( 1 ) age ≥ 18 years; ( 2 ) patients undergoing a single RIRS procedure without concurrent surgeries; ( 3 ) absence of specific conditions such as pregnancy, duplicated ureter, or horseshoe kidney; and ( 4 ) no prior ureteral stenting. This study was approved by the Institutional Review Board of KMUHIRB-F(I)-20200105 and SNUH1901-104-1005. Several patients and stone-related characteristics were used as input features for machine learning analysis, including age, sex, body mass index (BMI), stone laterality, number of stones, infundibular width (IW), pelvic stone angle (PSA) [12], average Hounsfield unit (HU) on computed tomography, stone burden, multifocality, and renal infundibular length (RIL). All patients underwent preoperative computed tomography (CT) to evaluate relevant anatomical parameters, including pelvic stone angle (PSA), renal infundibular length (RIL), and infundibular width (IW). Stone burden was quantified by calculating the cumulative stone diameter (CSD). The average Hounsfield Unit (HU) of each stone was determined by measuring attenuation values at both the center and outermost edges of the stone on CT imaging. For patients with multiple stones, the mean HU was calculated as the average of all stones identified. Multifocality was defined as the presence of stones in more than one renal calyx. Stone-free status (SFS) was defined as the absence of visible stones or residual fragments ≤ 5 mm on follow-up kidney-ureter-bladder (KUB) radiographs or CT scans obtained one month postoperatively. All imaging assessments were independently reviewed by two physicians. Model development, validation, and comparison The prediction model based on ML was developed using the dataset from KMUH. It was preprocessing and training the model with the available algorithms using a 10-fold cross-validation. Hyperparameter tuning was automatically performed via random grid search. KMUH dataset was split into the training dataset (n = 135) for model development and test dataset (n = 58) for internal validation by 7:3. Then, the model would be external validated using the SNUH dataset (n = 134) (Fig. 1 ). The accuracy (the proportion of all correct classification calculated by the formula: \(\:\:accuracy=\:\frac{\:SF\:correct\:classification+non\:SF\:correct\:classification}{all\:classification}\) ), AUC (area under the curve), Recall (sensitivity), Precision (Positive Predictive Value), F1-score (Harmonic mean of recall and precision: \(\:F1-score=\:\frac{2\times\:\:recall\times\:precision}{\:recall+precision}\) ), Kappa (Cohen’s kappa coefficient), MCC (Matthews Correlation Coefficient) are the used to compare the accuracy of model according to different ML algorithms, and then the accuracy and F1-score is the principle indications to determine the best model for further exploration. The feature importance plot is used demonstrate the ranking of feature importance using gain importance. SHAP (SHapley Additive exPlanations) is a game theory-based approach that can provide interpretability of ML models. It is another way to explain feature importance. The mean SHAP value of a feature indicates its contribution to the prediction. The positive or negative correlation between the feature and the predicted target depends on the mean SHAP value. In this study, features with the positive mean SHAP value mean that they have positive contribution to stone free (SF). On the contrary, features with the negative mean SHAP value negative related to SF, also can be considered positive related to non SF. Individual SHAP values in a violin plot are used to interpret the correlation, red points present Individuals with higher feature value, and Individuals with lower feature value show as blue points. When the red points are located to the right of the central axis, it indicates a positive correlation between the feature value and SHAP. Conversely, if the blue points are distributed on the right side of the central axis, it signifies a negative correlation between the feature value and SHAP. The mean absolute SHAP values are ranked and shown in a bar chart to demonstrate the importance ranking of features. Statistical analysis The associations between dependent and continuous variables were assessed using independent T-test (normal distribution) and Mann-Whitney U test (non-normal distribution), respectively. The assessment of association between the dependent and categorical variables was using Pearson's chi-square test or Fisher's exact test. Logistic regression was used to evaluate the relationship between SF and included features, and estimate the odds ratio. Python (version 3.11; Python Software Foundation) is the principal software to perform the analysis, and the ML was implemented with the PyCaret 3.0.4 package. PyCaret is an open-source, low-code ML learning library, including 16 ML algorithms for classification (Appendix A). Results The comparison of the training and test datasets is shown in Table 1 . All of the characteristics are no significant difference between the training [Mean(Q1-Q3) age: 60(50–67) years; Men: 67.36%] and the test [Mean(Q1-Q3) age: 58 (48–67) years; Men: 65.67%] datasets. Table 1 Comparison of patient characteristics between the training and test datasets Characteristics Training (n = 135) Test (n = 58) P RIL, mm 24.2(4.6) 24.5(4.7) 0.6628 IW, cm 0.54(0.40–0.77) 0.55(0.41–0.75) 0.8771 PSA 45.7(11.6) 47.6(14.4) 0.3737 HU 712(473–980) 838(547–1150) 0.1470 BMI, kg/m 2 26.7(24.1–29.1) 25.6(23.0-29.1) 0.2491 Age 60(50–67) 60.5(53–68) 0.2386 Stone number 2( 1 – 2 ) 1( 1 – 2 ) 0.5000 Stone burden, cm 1.65(0.95–2.54) 1.82(1.33–2.74) 0.2687 Gender, n (%) 0.9820 Women 44(32.60) 19(32.76) Men 91(67.40) 39(67.24) Multifocal, n (%) 0.3526 No 60(44.44) 30(51.72) Yes 75(55.56) 28(48.28) Stone side, n (%) 0.5781 Left 78(57.78) 31(53.44) Right 57(42.22) 27(46.56) DM 0.9642 No 81(60.00) 35(60.34) Yes 54(40.00) 23(39.66) HTN 0.4774 No 61(45.19) 23(39.66) Yes 74(54.81) 35(60.34) The results of model training are shown in Table 2 , LightGBM is the best classifier, with the highest accuracy in most indications except for Recall and AUC . In Fig. 2 a, there is slightly difference of accuracy between the prediction for non-SF and SF, the confused matrix presents the accuracy in non-SF is 54.54% (12/22) and in SF is 58.33% (21/36). The total accuracy in the test dataset is 56.90% (33/58). According to gini importance, the main features order is: PSA, HU, stone burden, BMI, RIL, IW, and age (Fig. 2 b). Table 2 The model performance for different ML algorithms Model Accuracy AUC Recall Prec. F1 Kappa MCC TT Light Gradient Boosting Machine 0.7181 0.7153 0.8194 0.7594 0.7812 0.3712 0.3931 0.0790 Random Forest Classifier 0.6967 0.7438 0.7972 0.7444 0.7641 0.3276 0.3440 0.0260 Logistic Regression 0.6731 0.7231 0.8083 0.7226 0.7514 0.2580 0.2900 0.0240 Naive Bayes 0.6654 0.7158 0.8097 0.7087 0.7482 0.2414 0.2607 0.0100 Quadratic Discriminant Analysis 0.6511 0.6603 0.7625 0.7162 0.7312 0.2248 0.2450 0.0090 Ridge Classifier 0.6429 0.7083 0.7847 0.6884 0.7241 0.1976 0.2145 0.0100 Decision Tree Classifier 0.639 0.609 0.7181 0.7240 0.7164 0.2105 0.2130 0.0100 Extra Trees Classifier 0.6374 0.7053 0.7764 0.6863 0.7252 0.1865 0.1990 0.0250 Linear Discriminant Analysis 0.6357 0.6992 0.7847 0.6833 0.7206 0.1739 0.1816 0.0100 Dummy Classifier 0.6291 0.5000 1.0000 0.6291 0.7723 0.0000 0.0000 0.0100 Ada Boost Classifier 0.6214 0.6103 0.7500 0.6805 0.7048 0.1578 0.1782 0.0160 Gradient Boosting Classifier 0.6209 0.6947 0.7375 0.6822 0.7044 0.1643 0.1736 0.0190 SVM - Linear Kernel 0.6198 0.5994 0.6972 0.7535 0.6430 0.1681 0.2015 0.0090 K Neighbors Classifier 0.6082 0.6329 0.7417 0.6742 0.7042 0.1217 0.1185 0.0140 AUC, area under the curve; F1, F1-score; κ, Cohen’s kappa coefficient; MCC, Matthews Correlation Coefficient; TT, training time. The prediction model performance in different datasets is summarized (Table 3 ). The validation in the SNUH dataset demonstrates better accuracy ( Accuracy : 0.7709, AUC : 0.7592, Recall : 0.8889, Precision : 0.8278, F1-score : 0.8534). For the all data (n = 327), the Accuracy is 0.6208, and the Recall and Precision are 0.7578 and 0.7026, respectively. Table 3 Comparison of the accuracy in different datasets Dataset Accuracy AUC Recall Precision F1-score Training (n = 135) 0.7181 0.7153 0.8194 0.7594 0.7812 Test (n = 58) 0.5690 0.5795 0.5833 0.6774 0.6269 KMUH (n = 193) 0.5745 0.5830 0.6853 0.6536 0.6657 SNUH(n = 134) 0.7709 0.7592 0.8889 0.8278 0.8534 ALL (n = 327) 0.6208 0.6096 0.7578 0.7026 0.7260 The SHAP analysis has highlighted the significance of features in the LightGBM model. As illustrated in Figs. 3 a and 3 b for KMUH and SNUH respectively, the violin plots indicate that higher values of stone burden, HU, RIL, and stone number correlate with a decreased prediction probability for SF. Conversely, individuals with elevated levels of PSA, age, DM, and HTN exhibit higher SHAP values. Those with either high or low BMI and IW demonstrate lower SHAP values compared to individuals with normal BMI and IW. The impact of gender, multifocality, and stone side on the model is minimal. These findings are consistent across both KMUH and SNUH. Figures 3 c and 3 d present all features for KMUH and SNUH, respectively, arranged by their mean absolute SHAP values. The stone burden is the most significant contributor to SF, followed by HU. The factors influencing the prediction model between KMUH and SNUH differ, especially in BMI and stone numbers. The importance of stone and patient characteristics for all participants in this study is illustrated in Fig. 4 , including both KMUH and SNUH. In addition to certain features with minimal contribution, such as DM, gender, multifocal, and stone side, there is a consistent relationship between features and SF in both logistic regression and machine learning analyses (see Table S2). Discussion Based on the guidelines published by the American Urological Association (AUA) and the European Association of Urology (EAU), percutaneous nephrolithotomy (PCNL) is recommended as the first-line surgical intervention for renal stone larger than 2 cm ( 13 ). In contrast, for renal stones measuring less than 2 cm, RIRS is generally considered the preferred approach. However, with ongoing advancements in fURS instruments and laser lithotripsy technologies, there has been a notable shift in clinical practice, with an increasing number of patients undergoing RIRS even for relatively large renal stones, instead of PCNL. RIRS has the advantages of fewer complications and faster postoperative recovery. Even though, the treatment of LPS remains inherently more challenging than stones in other renal locations. Anatomical constraints such as a steep IPA, elongated and narrow infundibula, and limited calyceal mobility hinder access and effective fragment clearance ( 14 ). Additionally, the presence of a laser fiber within the working channel compromises scope deflection, further complicating access to the lower pole. Giulioni C et al. large-scale study have demonstrated that larger stone burden, multiple stones, the use of reusable scopes are independently associated with residual fragments following RIRS ( 15 ). These findings emphasize the necessity of individualized surgical planning for LPS, including detailed preoperative anatomical assessment and careful selection of surgical strategies and instruments to optimize clinical outcomes. For accurate evaluation of the success of RIRS for LPS and prediction of SFR, several scoring systems were designed. The Resorlu-Unsal Stone Score (RUSS) scoring system incorporates several key parameters, including stone size, stone multiplicity, the presence of lower pole calculi with a renal IPA less than 45°, and abnormal renal anatomy such as horseshoe or pelvic kidneys ( 16 ). Another widely used scoring system is R.I.R.S scoring system which was summarized by Xiao et al. in 2017 ( 17 ). The system includes stone density, stone burden, renal infundibulopelvic length (RIL) and whether the stone is located in the lower calyx. As new technologies appear, the application of artificial intelligence (AI) including ML in endourology filed has dramatic development. From a previous research, it showed the increasing volume of related publications underscores a growing recognition of AI as a valuable tool in urolithiasis ( 18 ). Various AI models have been integrated to enhance accuracy in tasks ranging from predicting spontaneous stone passage, predicting SFR, diagnostic prediction, imaging interpretation, stone composition analysis, procedural outcome prediction and treatment planning ( 19 ). Kadlec et al. developed a predictive model aimed at prediction of SFR of various endourological procedures. It demonstrated a sensitivity of 75.3% and a specificity of 60.4% in predicting stone-free status, which was defined as the absence of stones on KUB X-ray or the presence of residual fragments less than 4 mm on CT. While the model exhibited high specificity (98.3%) in predicting the need for secondary interventions, its sensitivity in this regard was limited (30%). This investigation established a foundational framework for the future development of predictive nomograms in endourology ( 20 ). Carlotta Nedbal et al. described if ML could predict RIRS outcome including stone free status after surgery. It identified total stone burden as the most influential positive predictor of failure to achieve stone-free status (SFS), followed by the presence of a preoperative ureteral stent. Conversely, factors inversely associated with SFS failure included a negative preoperative urine culture and stone location, particularly when confined to a single calyx or the ureter which has been confirmed that LPS indeed affect the SFR ( 21 ). We demonstrate the potential of ML algorithms to predict SFR based on preoperative clinical parameters. In order to identify the highest accuracy model, we both conduct internal and external validation. Among the ML algorithms, the Light Gradient Boosting Machine had the best SFR prediction. The SHAP values highlight the relative importance impact of each variable on the ML model’s prediction of SFR. In both cohorts, stone burden emerged as the most influential factor, exhibiting the highest mean SHAP values, indicating that it remains the most significant variable associated with postoperative SFR outcomes. Other important contributors included Hounsfield units (HU), PSA, and RIL, indicating the relevance of kidney anatomical parameters in predicting outcomes. On the contrary, patient demographics such as age and BMI demonstrated moderate influence, it had limited predictive weight in both models. Moreover, the consistent patterns across both institutions suggest model generalizability and external validity. These findings underscore the value of integrating comprehensive preoperative data, particularly stone-related characteristics into AI-driven prediction models to enhance individualized surgical planning and optimize patient outcomes in endourology. From this research, it also validates that the PSA angle measurement proposed by our team is indeed associated with postoperative SFR, suggesting that it could serve as a reliable alternative to the traditional IPA measurement ( 12 ). Our study demonstrates the capability of ML algorithms to effectively handle complex and heterogeneous datasets, offering a significant advantage in predicting postoperative outcomes ( 22 ). By incorporating patient demographic information and preoperative clinical variables, ML models are able to generate accurate and individualized prognostic estimates. However, there are some limitations. First, it is a retrospective research and needs to confirm in prospective trials. Second, we did not include all possible related parameters which can impact on postoperative SFR. However, this study confirms the feasibility of using ML algorithms as a tool for predicting postoperative outcomes. It also paves the way for future applications of ML in large scale data analysis, which may further enhance predictive accuracy over time. While ML holds great potential in healthcare, its integration into postoperative care must be approached with caution, including concerns over data privacy, limited model interpretability, and the risk of algorithmic bias. Addressing these issues is essential to ensure the ethical and effective application of ML technologies in clinical practice. Conclusions This study demonstrates the feasibility and clinical utility of ML models in predicting SFR following RIRS for LPS. Among various algorithms, the Light Gradient Boosting Machine (LightGBM) achieved the highest predictive performance, with strong external validation in an independent cohort. The algorithms revealed that stone burden, Hounsfield unit, PSA which as a reliable alternative to the traditional IPA, and RIL were the most influential predictors of SFR, emphasizing the importance of both stone characteristics and renal anatomy in outcome prediction. These findings support the integration of ML-based decision support tools into preoperative planning for individualized surgical strategies. Future prospective studies with larger datasets and additional clinical variables are warranted to further refine and validate ML-based predictive models in endourology. Declarations Conflict of Interest The authors did not receive funding in any form from any organization for the submitted work. Author Contribution Conception and design: HYL, SYCAcquisition of data, analysis: YCW, YHTDrafting of the Manuscript: HYL, JCESupervision: VG Acknowledgements This study was partially supported by the Ministry of Science and Technology (MOST-111-2314-B-037-100-MY2, NSTC 113-2324-B-037-027), Kaohsiung Medical University Hospital (KMUH-111-1R55, KMUH-112-2R57, KMUH-113-3R49), and Regenerative Medicine and Cell Therapy Research Center (KMU-TC112A02). References Golomb D, Goldberg H, Tapiero S, Stabholz Y, Lotan P, Darawsha AE, Holland R, Ehrlich Y, Lifshitz D (2023) Retrograde intrarenal surgery for lower pole stones utilizing stone displacement technique yields excellent results. Asian J Urol 10(1):58–63 Dresner SL, Iremashvili V, Best SL, Hedican SP, Nakada SY (2020) Influence of lower pole infundibulopelvic angle on success of retrograde flexible ureteroscopy and laser lithotripsy for the treatment of renal stones. J Endourol 34(6):655–660 Yuri P, Hariwibowo R, Soeroharjo I, Danarto R, Hendri AZ, Brodjonegoro SR et al (2018) Meta-analysis of optimal management of lower pole stone of 10–20 mm: flexible ureteroscopy (FURS) versus extracorporeal shock wave lithotripsy (ESWL) versus percutaneous nephrolithotomy (PCNL). Acta Med Indones 50(1):18–25 Karim SS, Hanna L, Geraghty R, Somani BK (2020) Role of pelvicalyceal anatomy in the outcomes of retrograde intrarenal surgery (RIRS) for lower pole stones: outcomes with a systematic review of literature. Urolithiasis 48(3):263–270 Inoue T, Hamamoto S, Okada S, Imai S, Yamamichi F, Fujita M et al (2023) Pelvicalyceal anatomy on the accessibility of reusable flexible ureteroscopy to lower pole calyx during retrograde intrarenal surgery. Int J Urol 30(2):220–225 Schuster TG, Hollenbeck BK, Faerber GJ, Wolf JS Jr (2002) Ureteroscopic treatment of lower pole calculi: comparison of lithotripsy in situ and after displacement. J Urol 168(1):43–45 Ru Huang J-C, Chen Y-Q, Zhou J-J, Wang C-C, Hui M-J, Jiang et al (2024) Relocation of lower pole renal stones helps improve the stone-free rate during flexible ureteroscopy with a low complication rate. World J Urol 42(1):30 Kronenberg P, Traxer O (2019) The laser of the future: reality and expectations about the new thulium fiber laser-a systematic review. Transl Androl Urol 8(Suppl 4):S398–S417 Shrestha A, Traxer O, Seitz C, Corrales M, Castellani D, Chew BH et al (2024) Assessing flexible ureteroscopy outcomes for lower Pole versus non lower Pole stones using the flexible and navigable suction ureteric access sheath: a prospective multicenter study by EAU Endourology and PEARLS group. World J Urol 43(1):41 Dresner SL, Iremashvili V, Best SL, Hedican SP, Nakada SY (2020) Influence of Lower Pole Infundibulopelvic Angle on Success of Retrograde Flexible Ureteroscopy and Laser Lithotripsy for the Treatment of Renal Stones. J Endourol 34(6):655–660 Huang Y, Li K, Yang W, Li Z, Liu C, Lai C et al (2022) A Scoring System for Optimal Selection of Endoscopic Treatment for 1-2cm Lower Pole Renal Calculi. Urol J 19(5):356–362 Tung Y-H, Li W-M, Juan Y-S, Huang T-Y, Wang Y-C, Yeh H-C et al (2024) New infundibulopelvic angle measurement method can predict stone-free rates following retrograde intrarenal surgery. Sci Rep 14(1):9891 Awedew AF, Seman YS (2023) Dawit Zerihun Yalew, Yohannes Chemere Wondmeneh, Wassie Almaw Yigzaw. Efficacy and safety of surgical treatment for 1–2 cm sized lower pole of renal stone: network meta-analysis of randomized control trials. Urolithiasis 51(1):82 Anil Shrestha B, Adhikari F, Panthier S, Baidya V, Gauhar (2023) Olivier Traxer. Flexible ureteroscopy for lower pole calculus: is it still a challenge? World J Urol 41(11):3345–3353 Carlo Giulioni D, Castellani BK, Somani BH, Chew T, Tailly et al (2023) William Ong Lay Keat. The efficacy of retrograde intra-renal surgery (RIRS) for lower pole stones: results from 2946 patients. World J Urol. ;41(5):1407–1413 Resorlu B, Unsal A, Gulec H, Oztuna D (2012) A new scoring system for predicting stone-free rate after retrograde intrarenal surgery: The Resorlu-Unsal Stone Score. Urology 80(3):512–518 Xiao Y, Li D, Chen L, Xu Y, Zhang D, Shao Y, Lu J (2017) The R.I.R.S. scoring system: an innovative scoring system for predicting stone-free rate following retrograde intrarenal surgery. BMC Urol 17(1):1–8 Carlotta Nedbal CC 1, Jahrreiss V, Pietropaolo A, Galosi AB, Castellani D et al (2024) Trends of Artificial Intelligence, Machine Learning, Virtual Reality, and Radiomics in Urolithiasis over the Last 30 Years (1994–2023) as Published in the Literature (PubMed): A Comprehensive Review. J Endourol 38(8):788–798 Zeeshan Hameed BM, Shah M, Naik N, Rai BP, Karimi H, Rice P et al (2021) The Ascent of Artificial Intelligence in Endourology: a Systematic Review Over the Last 2 Decades. Curr Urol Rep 22(10):53 Kadlec A, Ohlander S, Hotaling J, Hannick J, Niederberger C, Turk TM (2014) Nonlinear logistic regression model for outcomes after endourologic procedures: a novel predictor. Urolithiasis 42:323–327 Carlotta Nedbal S, Adithya N, Naik S, Gite (2024) Patrick Juliebø-Jones, Bhaskar K Somani. Can Machine Learning Correctly Predict Outcomes of Flexible Ureteroscopy with Laser Lithotripsy for Kidney Stone Disease? Results from a Large Endourology University Centre. Eur Urol Open Sci 64:30–37 Elfanagely O, Toyoda Y, Othman S et al (2021) Machine learning and surgical outcomes prediction: a systematic review. J Surg Res 264:346–361 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 12 Jul, 2025 Read the published version in World Journal of Urology → Version 1 posted Editorial decision: Revision requested 13 May, 2025 Reviews received at journal 12 May, 2025 Reviewers agreed at journal 30 Apr, 2025 Reviewers agreed at journal 30 Apr, 2025 Reviewers invited by journal 30 Apr, 2025 Editor assigned by journal 30 Apr, 2025 Submission checks completed at journal 30 Apr, 2025 First submitted to journal 24 Apr, 2025 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-6517712","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":450953420,"identity":"8cae5403-4a37-46b3-90bc-db0a24cea311","order_by":0,"name":"Hsiang Ying Lee","email":"","orcid":"","institution":"Kaohsiung Medical University Chung-Ho Memorial Hospital","correspondingAuthor":false,"prefix":"","firstName":"Hsiang","middleName":"Ying","lastName":"Lee","suffix":""},{"id":450953421,"identity":"58ad4c82-1720-4740-a9d0-06cd3b2f1103","order_by":1,"name":"Sung Yong Cho","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAw0lEQVRIiWNgGAWjYBADORjDgDj1BxgYjEnXkthAtBbd9t6Hjz/uOZw+f3aPAcOPGgZj8wYCWszOHDc2OPDscO6GO2cMGHuOMZjJHCCk5UYam8SBA0AtEjkGDLwNDDYShBwG1ML+A6glXX5GjgHjXyK1sDEAtSQw3MgxYAbaYkZYy5ljzBJnDqQbbriRVnBY5piEMWEtx9sYP1QcsJaXn5G88eGbGhvDGYS0QEEzmDzAwEDQDjioI1rlKBgFo2AUjEAAAGhVP/9lkmdPAAAAAElFTkSuQmCC","orcid":"","institution":"Seoul National University Hospital","correspondingAuthor":true,"prefix":"","firstName":"Sung","middleName":"Yong","lastName":"Cho","suffix":""},{"id":450953422,"identity":"bf4beabf-84d2-4baa-bd87-a69c2ad9463a","order_by":2,"name":"Yu-Hung Tung","email":"","orcid":"","institution":"Kaohsiung Medical University Chung-Ho Memorial Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yu-Hung","middleName":"","lastName":"Tung","suffix":""},{"id":450953424,"identity":"b9ca0f77-a9f8-487c-a280-e9d8c754830b","order_by":3,"name":"Jose Carlo Elises","email":"","orcid":"","institution":"Seoul National University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Jose","middleName":"Carlo","lastName":"Elises","suffix":""},{"id":450953428,"identity":"5067a24b-6b81-44b3-85e4-1d863f55203b","order_by":4,"name":"Yen-Chun Wang","email":"","orcid":"","institution":"Kaohsiung Medical University Chung-Ho Memorial Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yen-Chun","middleName":"","lastName":"Wang","suffix":""},{"id":450953431,"identity":"eef8c6a3-d718-4311-9dad-82f570164224","order_by":5,"name":"Vineet Gauhar","email":"","orcid":"","institution":"Ng Teng Fong General Hospital, Asian Institute of Nephro-Urology","correspondingAuthor":false,"prefix":"","firstName":"Vineet","middleName":"","lastName":"Gauhar","suffix":""}],"badges":[],"createdAt":"2025-04-24 06:38:24","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6517712/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6517712/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s00345-025-05762-7","type":"published","date":"2025-07-12T15:57:52+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":82177379,"identity":"64fb495a-cbb0-4892-8232-e3e12c52d108","added_by":"auto","created_at":"2025-05-07 11:20:35","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":100683,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart of patient selection and dataset allocation for model development and validation.\u003c/p\u003e","description":"","filename":"Figure1Flowchart.png","url":"https://assets-eu.researchsquare.com/files/rs-6517712/v1/73990003e15c1c7a5425f424.png"},{"id":82178810,"identity":"3db1e8c2-e08b-43e3-862c-e05246b6f94c","added_by":"auto","created_at":"2025-05-07 11:28:35","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":33840,"visible":true,"origin":"","legend":"\u003cp\u003e(a) Confusion matrix of the Light Gradient Boosting Machine (LightGBM) classifier applied to the internal test dataset. The matrix displays true positive (21), true negative (12), false positive (10), and false negative (15) counts, corresponding to the prediction of stone-free status (1) or non-stone-free status (0) after retrograde intrarenal surgery (RIRS).\u003cbr\u003e\n(b) Feature importance plot ranked by gain-based variable importance from the LightGBM model.\u003c/p\u003e","description":"","filename":"Figure2Theconfusionmatrixandfeatureimportanceplot.png","url":"https://assets-eu.researchsquare.com/files/rs-6517712/v1/9ca0d2cd73311d1b5c8edb13.png"},{"id":82178811,"identity":"b5cce700-c4e2-486e-9966-6a3b77161dc5","added_by":"auto","created_at":"2025-05-07 11:28:35","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":102782,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP analysis of feature contributions to the prediction of stone-free status after retrograde intrarenal surgery (RIRS) in the LightGBM model. Violin plots (a, c) illustrate the distribution and impact of each feature's SHAP value, with red indicating higher feature values and blue indicating lower values. Positive SHAP values are associated with higher predicted probability of stone-free outcome. Bar plots (b, d) rank features based on their mean absolute SHAP values, reflecting overall importance in model prediction.\u003c/p\u003e","description":"","filename":"Figure3.TheSHAPanalysisofallfeaturesforKMUHandSNUH..png","url":"https://assets-eu.researchsquare.com/files/rs-6517712/v1/84f0fd971191481dbe74f5cb.png"},{"id":82179390,"identity":"a3815bd4-ad01-46ae-b92b-5b07196262e5","added_by":"auto","created_at":"2025-05-07 11:36:35","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":94788,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP analysis of feature contribution in the combined dataset (KMUH + SNUH). (a) The violin plot, (b) The bar chart.\u003c/p\u003e","description":"","filename":"Figure4.TheSHAPanalysisofallfeaturesforallcases..png","url":"https://assets-eu.researchsquare.com/files/rs-6517712/v1/2ad1a413bbb76fd6e1100b11.png"},{"id":86699445,"identity":"ead2326f-df26-4280-93e3-a599abd15c20","added_by":"auto","created_at":"2025-07-14 16:09:45","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1022980,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6517712/v1/41fab8c4-3ea4-478e-919f-b7fac211a55e.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Machine Learning-Based Prediction of Stone-Free Rate After Retrograde Intrarenal Surgery for Lower Pole Renal Stones","fulltext":[{"header":"Introduction","content":"\u003cp\u003eLower pole stones (LPS) is the most common position of renal stones, and it comprised about 25%~35% of renal stones (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). LPS were considered as challenging position for retrograde intrarenal surgery (RIRS) due to the unique anatomical characteristics with difficulty in removing fragments. It can limit access for lithotripsy and make it difficult to extract stone fragments, even after comprehensive treatment (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). Although laser technology, endoscopic instruments, and lithotripsy accessories have seen significant advancements in recent years, the stone-free rate (SFR) for LPS remains lower compared to stones in other locations (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe SFR for LPS treated with flexible ureteroscopy (fURS) remains suboptimal. One major factor is the presence of an acute infundibulopelvic angle (IPA) and a long, narrow infundibulum, both of which hinder access to the lower pole calyces (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). Although advancements in ureteroscope technology have improved deflection capabilities, the insertion of instruments such as laser fibers or stone retrieval baskets through the working channel can still significantly limit scope maneuverability (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). To address these challenges and improve SFR, several techniques have been proposed. These include stone relocation from the lower pole to more accessible calyces (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e), the use of smaller diameter laser fibers or baskets to enhance deflection, and the application of dusting modes that facilitate spontaneous clearance of stone fragments (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e). Additionally, the development and adoption of suction-assisted access sheaths have shown promise in fragment evacuation (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e). Despite these innovations, the implementation of targeted strategies specifically designed to optimize the management of LPS remains essential for improving surgical outcomes.\u003c/p\u003e \u003cp\u003eFrom previous literature, preoperative predictive scoring system or parameters were designed to predict the SFR of LPS. Stephanie L. Dresner et al. (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e) demonstrated the role of IPA. The more acute IPA and larger preoperative stone size negatively affect SFR and need for repeat surgery. Yuleng Huang et al. (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e) propose a scoring system comprised of stone characteristics, collecting system anatomy to predict the SFR of 1\u0026ndash;2 cm LPS. Due to the difficulty of position and better decision-making before treatment plan, we conducted a machine-learning (ML) prediction model including different aspects of characteristics for better prediction.\u003c/p\u003e \u003cp\u003eMethods and materials\u003c/p\u003e \u003cp\u003eWe conducted a retrospective analysis of 327 patients with lower pole renal stones who underwent RIRS. Among them, 193 patients were treated at Kaohsiung Medical University Hospital (KMUH) between April 2018 and October 2023, and 134 patients were treated at Seoul National University Hospital (SNUH) from January 2022 to January 2024. Patients with rare or atypical conditions were excluded, as their limited numbers could have introduced bias and affected the generalizability of the results. Inclusion criteria were as follows: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) age\u0026thinsp;\u0026ge;\u0026thinsp;18 years; (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) patients undergoing a single RIRS procedure without concurrent surgeries; (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) absence of specific conditions such as pregnancy, duplicated ureter, or horseshoe kidney; and (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) no prior ureteral stenting. This study was approved by the Institutional Review Board of KMUHIRB-F(I)-20200105 and SNUH1901-104-1005. Several patients and stone-related characteristics were used as input features for machine learning analysis, including age, sex, body mass index (BMI), stone laterality, number of stones, infundibular width (IW), pelvic stone angle (PSA) [12], average Hounsfield unit (HU) on computed tomography, stone burden, multifocality, and renal infundibular length (RIL).\u003c/p\u003e \u003cp\u003eAll patients underwent preoperative computed tomography (CT) to evaluate relevant anatomical parameters, including pelvic stone angle (PSA), renal infundibular length (RIL), and infundibular width (IW). Stone burden was quantified by calculating the cumulative stone diameter (CSD). The average Hounsfield Unit (HU) of each stone was determined by measuring attenuation values at both the center and outermost edges of the stone on CT imaging. For patients with multiple stones, the mean HU was calculated as the average of all stones identified. Multifocality was defined as the presence of stones in more than one renal calyx. Stone-free status (SFS) was defined as the absence of visible stones or residual fragments\u0026thinsp;\u0026le;\u0026thinsp;5 mm on follow-up kidney-ureter-bladder (KUB) radiographs or CT scans obtained one month postoperatively. All imaging assessments were independently reviewed by two physicians.\u003c/p\u003e"},{"header":"Model development, validation, and comparison","content":"\u003cp\u003eThe prediction model based on ML was developed using the dataset from KMUH. It was preprocessing and training the model with the available algorithms using a 10-fold cross-validation. Hyperparameter tuning was automatically performed via random grid search. KMUH dataset was split into the training dataset (n\u0026thinsp;=\u0026thinsp;135) for model development and test dataset (n\u0026thinsp;=\u0026thinsp;58) for internal validation by 7:3. Then, the model would be external validated using the SNUH dataset (n\u0026thinsp;=\u0026thinsp;134) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe \u003cb\u003eaccuracy\u003c/b\u003e (the proportion of all correct classification calculated by the formula:\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\:accuracy=\\:\\frac{\\:SF\\:correct\\:classification+non\\:SF\\:correct\\:classification}{all\\:classification}\\)\u003c/span\u003e\u003c/span\u003e), \u003cb\u003eAUC\u003c/b\u003e (area under the curve), \u003cb\u003eRecall\u003c/b\u003e (sensitivity), \u003cb\u003ePrecision\u003c/b\u003e (Positive Predictive Value), \u003cb\u003eF1-score\u003c/b\u003e (Harmonic mean of recall and precision:\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:F1-score=\\:\\frac{2\\times\\:\\:recall\\times\\:precision}{\\:recall+precision}\\)\u003c/span\u003e\u003c/span\u003e), \u003cb\u003eKappa\u003c/b\u003e (Cohen\u0026rsquo;s kappa coefficient), \u003cb\u003eMCC\u003c/b\u003e (Matthews Correlation Coefficient) are the used to compare the accuracy of model according to different ML algorithms, and then the \u003cb\u003eaccuracy\u003c/b\u003e and \u003cb\u003eF1-score\u003c/b\u003e is the principle indications to determine the best model for further exploration. The feature importance plot is used demonstrate the ranking of feature importance using gain importance.\u003c/p\u003e \u003cp\u003eSHAP (SHapley Additive exPlanations) is a game theory-based approach that can provide interpretability of ML models. It is another way to explain feature importance. The mean SHAP value of a feature indicates its contribution to the prediction. The positive or negative correlation between the feature and the predicted target depends on the mean SHAP value. In this study, features with the positive mean SHAP value mean that they have positive contribution to stone free (SF). On the \u003cem\u003econtrary, features with the negative mean SHAP value negative related to SF, also\u003c/em\u003e can be considered positive related to non SF.\u003c/p\u003e \u003cp\u003eIndividual SHAP values in a violin plot are used to interpret the correlation, red points present Individuals with higher feature value, and Individuals with lower feature value show as blue points. When the red points are located to the right of the central axis, it indicates a positive correlation between the feature value and SHAP. Conversely, if the blue points are distributed on the right side of the central axis, it signifies a negative correlation between the feature value and SHAP. The mean absolute SHAP values are ranked and shown in a bar chart to demonstrate the importance ranking of features.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eThe associations between dependent and continuous variables were assessed using independent T-test (normal distribution) and Mann-Whitney U test (non-normal distribution), respectively. The assessment of association between the dependent and categorical variables was using Pearson's chi-square test or Fisher's exact test. Logistic regression was used to evaluate the relationship between SF and included features, and estimate the odds ratio.\u003c/p\u003e \u003cp\u003ePython (version 3.11; Python Software Foundation) is the principal software to perform the analysis, and the ML was implemented with the PyCaret 3.0.4 package. PyCaret is an open-source, low-code ML learning library, including 16 ML algorithms for classification (Appendix A).\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003eThe comparison of the training and test datasets is shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. All of the characteristics are no significant difference between the training [Mean(Q1-Q3) age: 60(50\u0026ndash;67) years; Men: 67.36%] and the test [Mean(Q1-Q3) age: 58 (48\u0026ndash;67) years; Men: 65.67%] datasets.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eComparison of patient characteristics between the training and test datasets\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCharacteristics\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTraining (n\u0026thinsp;=\u0026thinsp;135)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTest (n\u0026thinsp;=\u0026thinsp;58)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eP\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRIL, mm\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e24.2(4.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e24.5(4.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.6628\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIW, cm\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.54(0.40\u0026ndash;0.77)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.55(0.41\u0026ndash;0.75)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8771\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePSA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e45.7(11.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e47.6(14.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.3737\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHU\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e712(473\u0026ndash;980)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e838(547\u0026ndash;1150)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.1470\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBMI, kg/m\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e26.7(24.1\u0026ndash;29.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e25.6(23.0-29.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.2491\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e60(50\u0026ndash;67)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e60.5(53\u0026ndash;68)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.2386\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStone number\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.5000\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStone burden, cm\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.65(0.95\u0026ndash;2.54)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.82(1.33\u0026ndash;2.74)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.2687\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGender, n (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.9820\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWomen\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e44(32.60)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e19(32.76)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMen\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e91(67.40)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e39(67.24)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMultifocal, n (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.3526\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e60(44.44)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e30(51.72)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e75(55.56)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e28(48.28)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStone side, n (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.5781\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLeft\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e78(57.78)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e31(53.44)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRight\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e57(42.22)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e27(46.56)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.9642\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e81(60.00)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e35(60.34)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e54(40.00)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e23(39.66)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHTN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.4774\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e61(45.19)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e23(39.66)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e74(54.81)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e35(60.34)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe results of model training are shown in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, LightGBM is the best classifier, with the highest accuracy in most indications except for \u003cb\u003eRecall\u003c/b\u003e and \u003cb\u003eAUC\u003c/b\u003e. In Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e2\u003c/span\u003ea, there is slightly difference of accuracy between the prediction for non-SF and SF, the confused matrix presents the accuracy in non-SF is 54.54% (12/22) and in SF is 58.33% (21/36). The total accuracy in the test dataset is 56.90% (33/58). According to gini importance, the main features order is: PSA, HU, stone burden, BMI, RIL, IW, and age (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e2\u003c/span\u003eb).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe model performance for different ML algorithms\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAUC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRecall\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003ePrec.\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eF1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eKappa\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eMCC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eTT\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLight Gradient Boosting Machine\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.7181\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.7153\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8194\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.7594\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7812\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.3712\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.3931\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.0790\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRandom Forest Classifier\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.6967\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.7438\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.7972\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.7444\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7641\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.3276\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.3440\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.0260\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLogistic Regression\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.6731\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.7231\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8083\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.7226\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7514\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.2580\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.2900\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.0240\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNaive Bayes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.6654\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.7158\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8097\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.7087\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7482\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.2414\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.2607\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.0100\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eQuadratic Discriminant Analysis\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.6511\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.6603\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.7625\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.7162\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7312\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.2248\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.2450\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.0090\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRidge Classifier\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.6429\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.7083\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.7847\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.6884\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7241\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.1976\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.2145\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.0100\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDecision Tree Classifier\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.639\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.609\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.7181\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.7240\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7164\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.2105\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.2130\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.0100\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eExtra Trees Classifier\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.6374\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.7053\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.7764\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.6863\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7252\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.1865\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.1990\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.0250\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLinear Discriminant Analysis\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.6357\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.6992\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.7847\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.6833\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7206\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.1739\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.1816\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.0100\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDummy Classifier\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.6291\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.5000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.6291\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7723\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.0100\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAda Boost Classifier\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.6214\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.6103\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.7500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.6805\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7048\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.1578\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.1782\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.0160\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGradient Boosting Classifier\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.6209\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.6947\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.7375\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.6822\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7044\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.1643\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.1736\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.0190\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSVM - Linear Kernel\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.6198\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.5994\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.6972\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.7535\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.6430\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.1681\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.2015\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.0090\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eK Neighbors Classifier\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.6082\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.6329\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.7417\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.6742\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7042\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.1217\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.1185\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.0140\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"9\"\u003eAUC, area under the curve; F1, F1-score; κ, Cohen\u0026rsquo;s kappa coefficient; MCC, Matthews Correlation Coefficient; TT, training time.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe prediction model performance in different datasets is summarized (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). The validation in the SNUH dataset demonstrates better accuracy (\u003cb\u003eAccuracy\u003c/b\u003e: 0.7709, \u003cb\u003eAUC\u003c/b\u003e: 0.7592, \u003cb\u003eRecall\u003c/b\u003e: 0.8889, \u003cb\u003ePrecision\u003c/b\u003e: 0.8278, \u003cb\u003eF1-score\u003c/b\u003e: 0.8534). For the all data (n\u0026thinsp;=\u0026thinsp;327), the \u003cb\u003eAccuracy\u003c/b\u003e is 0.6208, and the \u003cb\u003eRecall\u003c/b\u003e and \u003cb\u003ePrecision\u003c/b\u003e are 0.7578 and 0.7026, respectively.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eComparison of the accuracy in different datasets\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDataset\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAUC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRecall\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003ePrecision\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eF1-score\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTraining (n\u0026thinsp;=\u0026thinsp;135)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.7181\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.7153\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8194\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.7594\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7812\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTest (n\u0026thinsp;=\u0026thinsp;58)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.5690\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.5795\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.5833\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.6774\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.6269\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKMUH (n\u0026thinsp;=\u0026thinsp;193)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.5745\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.5830\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.6853\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.6536\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.6657\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSNUH(n\u0026thinsp;=\u0026thinsp;134)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.7709\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.7592\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8889\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.8278\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.8534\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eALL (n\u0026thinsp;=\u0026thinsp;327)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.6208\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.6096\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.7578\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.7026\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7260\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe SHAP analysis has highlighted the significance of features in the LightGBM model. As illustrated in Figs.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e3\u003c/span\u003ea and \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e3\u003c/span\u003eb for KMUH and SNUH respectively, the violin plots indicate that higher values of stone burden, HU, RIL, and stone number correlate with a decreased prediction probability for SF. Conversely, individuals with elevated levels of PSA, age, DM, and HTN exhibit higher SHAP values. Those with either high or low BMI and IW demonstrate lower SHAP values compared to individuals with normal BMI and IW. The impact of gender, multifocality, and stone side on the model is minimal. These findings are consistent across both KMUH and SNUH.\u003c/p\u003e \u003cp\u003eFigures \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e3\u003c/span\u003ec and \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e3\u003c/span\u003ed present all features for KMUH and SNUH, respectively, arranged by their mean absolute SHAP values. The stone burden is the most significant contributor to SF, followed by HU. The factors influencing the prediction model between KMUH and SNUH differ, especially in BMI and stone numbers.\u003c/p\u003e \u003cp\u003eThe importance of stone and patient characteristics for all participants in this study is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, including both KMUH and SNUH. In addition to certain features with minimal contribution, such as DM, gender, multifocal, and stone side, there is a consistent relationship between features and SF in both logistic regression and machine learning analyses (see Table S2).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eBased on the guidelines published by the American Urological Association (AUA) and the European Association of Urology (EAU), percutaneous nephrolithotomy (PCNL) is recommended as the first-line surgical intervention for renal stone larger than 2 cm (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e). In contrast, for renal stones measuring less than 2 cm, RIRS is generally considered the preferred approach. However, with ongoing advancements in fURS instruments and laser lithotripsy technologies, there has been a notable shift in clinical practice, with an increasing number of patients undergoing RIRS even for relatively large renal stones, instead of PCNL. RIRS has the advantages of fewer complications and faster postoperative recovery.\u003c/p\u003e \u003cp\u003eEven though, the treatment of LPS remains inherently more challenging than stones in other renal locations. Anatomical constraints such as a steep IPA, elongated and narrow infundibula, and limited calyceal mobility hinder access and effective fragment clearance (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). Additionally, the presence of a laser fiber within the working channel compromises scope deflection, further complicating access to the lower pole. Giulioni C et al. large-scale study have demonstrated that larger stone burden, multiple stones, the use of reusable scopes are independently associated with residual fragments following RIRS (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e). These findings emphasize the necessity of individualized surgical planning for LPS, including detailed preoperative anatomical assessment and careful selection of surgical strategies and instruments to optimize clinical outcomes.\u003c/p\u003e \u003cp\u003eFor accurate evaluation of the success of RIRS for LPS and prediction of SFR, several scoring systems were designed. The Resorlu-Unsal Stone Score (RUSS) scoring system incorporates several key parameters, including stone size, stone multiplicity, the presence of lower pole calculi with a renal IPA less than 45\u0026deg;, and abnormal renal anatomy such as horseshoe or pelvic kidneys (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e). Another widely used scoring system is R.I.R.S scoring system which was summarized by Xiao et al. in 2017 (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e). The system includes stone density, stone burden, renal infundibulopelvic length (RIL) and whether the stone is located in the lower calyx. As new technologies appear, the application of artificial intelligence (AI) including ML in endourology filed has dramatic development. From a previous research, it showed the increasing volume of related publications underscores a growing recognition of AI as a valuable tool in urolithiasis (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eVarious AI models have been integrated to enhance accuracy in tasks ranging from predicting spontaneous stone passage, predicting SFR, diagnostic prediction, imaging interpretation, stone composition analysis, procedural outcome prediction and treatment planning (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e). Kadlec et al. developed a predictive model aimed at prediction of SFR of various endourological procedures. It demonstrated a sensitivity of 75.3% and a specificity of 60.4% in predicting stone-free status, which was defined as the absence of stones on KUB X-ray or the presence of residual fragments less than 4 mm on CT. While the model exhibited high specificity (98.3%) in predicting the need for secondary interventions, its sensitivity in this regard was limited (30%). This investigation established a foundational framework for the future development of predictive nomograms in endourology (\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eCarlotta Nedbal et al. described if ML could predict RIRS outcome including stone free status after surgery. It identified total stone burden as the most influential positive predictor of failure to achieve stone-free status (SFS), followed by the presence of a preoperative ureteral stent. Conversely, factors inversely associated with SFS failure included a negative preoperative urine culture and stone location, particularly when confined to a single calyx or the ureter which has been confirmed that LPS indeed affect the SFR (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e). We demonstrate the potential of ML algorithms to predict SFR based on preoperative clinical parameters.\u003c/p\u003e \u003cp\u003eIn order to identify the highest accuracy model, we both conduct internal and external validation. Among the ML algorithms, the Light Gradient Boosting Machine had the best SFR prediction. The SHAP values highlight the relative importance impact of each variable on the ML model\u0026rsquo;s prediction of SFR. In both cohorts, stone burden emerged as the most influential factor, exhibiting the highest mean SHAP values, indicating that it remains the most significant variable associated with postoperative SFR outcomes. Other important contributors included Hounsfield units (HU), PSA, and RIL, indicating the relevance of kidney anatomical parameters in predicting outcomes. On the contrary, patient demographics such as age and BMI demonstrated moderate influence, it had limited predictive weight in both models. Moreover, the consistent patterns across both institutions suggest model generalizability and external validity. These findings underscore the value of integrating comprehensive preoperative data, particularly stone-related characteristics into AI-driven prediction models to enhance individualized surgical planning and optimize patient outcomes in endourology. From this research, it also validates that the PSA angle measurement proposed by our team is indeed associated with postoperative SFR, suggesting that it could serve as a reliable alternative to the traditional IPA measurement (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eOur study demonstrates the capability of ML algorithms to effectively handle complex and heterogeneous datasets, offering a significant advantage in predicting postoperative outcomes (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e). By incorporating patient demographic information and preoperative clinical variables, ML models are able to generate accurate and individualized prognostic estimates. However, there are some limitations. First, it is a retrospective research and needs to confirm in prospective trials. Second, we did not include all possible related parameters which can impact on postoperative SFR. However, this study confirms the feasibility of using ML algorithms as a tool for predicting postoperative outcomes. It also paves the way for future applications of ML in large scale data analysis, which may further enhance predictive accuracy over time. While ML holds great potential in healthcare, its integration into postoperative care must be approached with caution, including concerns over data privacy, limited model interpretability, and the risk of algorithmic bias. Addressing these issues is essential to ensure the ethical and effective application of ML technologies in clinical practice.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eThis study demonstrates the feasibility and clinical utility of ML models in predicting SFR following RIRS for LPS. Among various algorithms, the Light Gradient Boosting Machine (LightGBM) achieved the highest predictive performance, with strong external validation in an independent cohort. The algorithms revealed that stone burden, Hounsfield unit, PSA which as a reliable alternative to the traditional IPA, and RIL were the most influential predictors of SFR, emphasizing the importance of both stone characteristics and renal anatomy in outcome prediction. These findings support the integration of ML-based decision support tools into preoperative planning for individualized surgical strategies. Future prospective studies with larger datasets and additional clinical variables are warranted to further refine and validate ML-based predictive models in endourology.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eConflict of Interest\u003c/h2\u003e \u003cp\u003eThe authors did not receive funding in any form from any organization for the submitted work.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eConception and design: HYL, SYCAcquisition of data, analysis: YCW, YHTDrafting of the Manuscript: HYL, JCESupervision: VG\u003c/p\u003e\u003ch2\u003eAcknowledgements\u003c/h2\u003e \u003cp\u003eThis study was partially supported by the Ministry of Science and Technology (MOST-111-2314-B-037-100-MY2, NSTC 113-2324-B-037-027), Kaohsiung Medical University Hospital (KMUH-111-1R55, KMUH-112-2R57, KMUH-113-3R49), and Regenerative Medicine and Cell Therapy Research Center (KMU-TC112A02).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eGolomb D, Goldberg H, Tapiero S, Stabholz Y, Lotan P, Darawsha AE, Holland R, Ehrlich Y, Lifshitz D (2023) Retrograde intrarenal surgery for lower pole stones utilizing stone displacement technique yields excellent results. Asian J Urol 10(1):58\u0026ndash;63\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDresner SL, Iremashvili V, Best SL, Hedican SP, Nakada SY (2020) Influence of lower pole infundibulopelvic angle on success of retrograde flexible ureteroscopy and laser lithotripsy for the treatment of renal stones. J Endourol 34(6):655\u0026ndash;660\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYuri P, Hariwibowo R, Soeroharjo I, Danarto R, Hendri AZ, Brodjonegoro SR et al (2018) Meta-analysis of optimal management of lower pole stone of 10\u0026ndash;20 mm: flexible ureteroscopy (FURS) versus extracorporeal shock wave lithotripsy (ESWL) versus percutaneous nephrolithotomy (PCNL). Acta Med Indones 50(1):18\u0026ndash;25\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKarim SS, Hanna L, Geraghty R, Somani BK (2020) Role of pelvicalyceal anatomy in the outcomes of retrograde intrarenal surgery (RIRS) for lower pole stones: outcomes with a systematic review of literature. Urolithiasis 48(3):263\u0026ndash;270\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eInoue T, Hamamoto S, Okada S, Imai S, Yamamichi F, Fujita M et al (2023) Pelvicalyceal anatomy on the accessibility of reusable flexible ureteroscopy to lower pole calyx during retrograde intrarenal surgery. Int J Urol 30(2):220\u0026ndash;225\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchuster TG, Hollenbeck BK, Faerber GJ, Wolf JS Jr (2002) Ureteroscopic treatment of lower pole calculi: comparison of lithotripsy in situ and after displacement. J Urol 168(1):43\u0026ndash;45\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRu Huang J-C, Chen Y-Q, Zhou J-J, Wang C-C, Hui M-J, Jiang et al (2024) Relocation of lower pole renal stones helps improve the stone-free rate during flexible ureteroscopy with a low complication rate. World J Urol 42(1):30\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKronenberg P, Traxer O (2019) The laser of the future: reality and expectations about the new thulium fiber laser-a systematic review. Transl Androl Urol 8(Suppl 4):S398\u0026ndash;S417\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShrestha A, Traxer O, Seitz C, Corrales M, Castellani D, Chew BH et al (2024) Assessing flexible ureteroscopy outcomes for lower Pole versus non lower Pole stones using the flexible and navigable suction ureteric access sheath: a prospective multicenter study by EAU Endourology and PEARLS group. World J Urol 43(1):41\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDresner SL, Iremashvili V, Best SL, Hedican SP, Nakada SY (2020) Influence of Lower Pole Infundibulopelvic Angle on Success of Retrograde Flexible Ureteroscopy and Laser Lithotripsy for the Treatment of Renal Stones. J Endourol 34(6):655\u0026ndash;660\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang Y, Li K, Yang W, Li Z, Liu C, Lai C et al (2022) A Scoring System for Optimal Selection of Endoscopic Treatment for 1-2cm Lower Pole Renal Calculi. Urol J 19(5):356\u0026ndash;362\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTung Y-H, Li W-M, Juan Y-S, Huang T-Y, Wang Y-C, Yeh H-C et al (2024) New infundibulopelvic angle measurement method can predict stone-free rates following retrograde intrarenal surgery. Sci Rep 14(1):9891\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAwedew AF, Seman YS (2023) Dawit Zerihun Yalew, Yohannes Chemere Wondmeneh, Wassie Almaw Yigzaw. Efficacy and safety of surgical treatment for 1\u0026ndash;2 cm sized lower pole of renal stone: network meta-analysis of randomized control trials. Urolithiasis 51(1):82\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnil Shrestha B, Adhikari F, Panthier S, Baidya V, Gauhar (2023) Olivier Traxer. Flexible ureteroscopy for lower pole calculus: is it still a challenge? World J Urol 41(11):3345\u0026ndash;3353\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarlo Giulioni D, Castellani BK, Somani BH, Chew T, Tailly et al (2023) William Ong Lay Keat. The efficacy of retrograde intra-renal surgery (RIRS) for lower pole stones: results from 2946 patients. World J Urol. ;41(5):1407\u0026ndash;1413\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eResorlu B, Unsal A, Gulec H, Oztuna D (2012) A new scoring system for predicting stone-free rate after retrograde intrarenal surgery: The Resorlu-Unsal Stone Score. Urology 80(3):512\u0026ndash;518\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXiao Y, Li D, Chen L, Xu Y, Zhang D, Shao Y, Lu J (2017) The R.I.R.S. scoring system: an innovative scoring system for predicting stone-free rate following retrograde intrarenal surgery. BMC Urol 17(1):1\u0026ndash;8\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarlotta Nedbal CC 1, Jahrreiss V, Pietropaolo A, Galosi AB, Castellani D et al (2024) Trends of Artificial Intelligence, Machine Learning, Virtual Reality, and Radiomics in Urolithiasis over the Last 30 Years (1994\u0026ndash;2023) as Published in the Literature (PubMed): A Comprehensive Review. J Endourol 38(8):788\u0026ndash;798\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZeeshan Hameed BM, Shah M, Naik N, Rai BP, Karimi H, Rice P et al (2021) The Ascent of Artificial Intelligence in Endourology: a Systematic Review Over the Last 2 Decades. Curr Urol Rep 22(10):53\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKadlec A, Ohlander S, Hotaling J, Hannick J, Niederberger C, Turk TM (2014) Nonlinear logistic regression model for outcomes after endourologic procedures: a novel predictor. Urolithiasis 42:323\u0026ndash;327\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarlotta Nedbal S, Adithya N, Naik S, Gite (2024) Patrick Julieb\u0026oslash;-Jones, Bhaskar K Somani. Can Machine Learning Correctly Predict Outcomes of Flexible Ureteroscopy with Laser Lithotripsy for Kidney Stone Disease? Results from a Large Endourology University Centre. Eur Urol Open Sci 64:30\u0026ndash;37\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eElfanagely O, Toyoda Y, Othman S et al (2021) Machine learning and surgical outcomes prediction: a systematic review. J Surg Res 264:346\u0026ndash;361\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"world-journal-of-urology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"wjur","sideBox":"Learn more about [World Journal of Urology](https://link.springer.com/journal/345)","snPcode":"345","submissionUrl":"https://submission.nature.com/new-submission/345/3","title":"World Journal of Urology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6517712/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6517712/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground:\u003c/h2\u003e \u003cp\u003eLower pole renal stones (LPS) present unique challenges for retrograde intrarenal surgery (RIRS) due to unfavorable anatomical features, often resulting in suboptimal stone-free rates (SFR). Recent advancements in machine learning (ML) offer new opportunities to predict surgical outcomes and guide clinical decision-making. This study aimed to develop and validate ML-based models to predict SFR following RIRS for LPS.\u003c/p\u003e\u003ch2\u003eMaterials and Methods:\u003c/h2\u003e \u003cp\u003eWe retrospectively analyzed data from 327 patients with LPS who underwent RIRS at two academic institutions: Kaohsiung Medical University Hospital (KMUH, n\u0026thinsp;=\u0026thinsp;193) and Seoul National University Hospital (SNUH, n\u0026thinsp;=\u0026thinsp;134). Demographic, anatomical, and stone-related variables were collected, including stone burden, Hounsfield unit (HU), pelvic stone angle (PSA), and renal infundibular length (RIL). A Light Gradient Boosting Machine (LightGBM) algorithm was developed using KMUH data and externally validated with SNUH data. SHAP (SHapley Additive exPlanations) analysis was performed to interpret feature importance.\u003c/p\u003e\u003ch2\u003eResults:\u003c/h2\u003e \u003cp\u003eThe LightGBM model achieved the highest predictive performance. External validation using the SNUH dataset yielded an accuracy of 77.1%, AUC of 0.759, and F1-score of 0.853. SHAP analysis revealed that stone burden, HU, PSA, and RIL were the most influential features. Notably, PSA demonstrated strong predictive relevance, supporting its use as an alternative to the traditional infundibulopelvic angle (IPA) in anatomical assessment.\u003c/p\u003e\u003ch2\u003eConclusions:\u003c/h2\u003e \u003cp\u003eML-based models, particularly LightGBM, offer robust predictive capability for SFR following RIRS in patients with LPS. These tools may enhance preoperative planning and personalized surgical strategies. Future prospective studies are warranted to further validate their clinical utility and expand on feature integration.\u003c/p\u003e","manuscriptTitle":"Machine Learning-Based Prediction of Stone-Free Rate After Retrograde Intrarenal Surgery for Lower Pole Renal Stones","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-07 11:20:30","doi":"10.21203/rs.3.rs-6517712/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-05-13T11:26:56+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-05-12T16:02:26+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"92831462984667338998411647050176802331","date":"2025-04-30T13:01:09+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"49776530374144957702782018965203339283","date":"2025-04-30T12:06:39+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-04-30T11:52:53+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-04-30T11:51:43+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-04-30T11:18:13+00:00","index":"","fulltext":""},{"type":"submitted","content":"World Journal of Urology","date":"2025-04-24T06:32:49+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"world-journal-of-urology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"wjur","sideBox":"Learn more about [World Journal of Urology](https://link.springer.com/journal/345)","snPcode":"345","submissionUrl":"https://submission.nature.com/new-submission/345/3","title":"World Journal of Urology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"f1bceaba-56fc-4d0a-8736-8afba8dffeee","owner":[],"postedDate":"May 7th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-07-14T16:03:55+00:00","versionOfRecord":{"articleIdentity":"rs-6517712","link":"https://doi.org/10.1007/s00345-025-05762-7","journal":{"identity":"world-journal-of-urology","isVorOnly":false,"title":"World Journal of Urology"},"publishedOn":"2025-07-12 15:57:52","publishedOnDateReadable":"July 12th, 2025"},"versionCreatedAt":"2025-05-07 11:20:30","video":"","vorDoi":"10.1007/s00345-025-05762-7","vorDoiUrl":"https://doi.org/10.1007/s00345-025-05762-7","workflowStages":[]},"version":"v1","identity":"rs-6517712","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6517712","identity":"rs-6517712","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}