Construction method of machine deep learning prediction model for urinary tract infection risk after pediatric pyeloplasty, and related system and equipment

Abstract

The invention relates to a construction method of a machine deep learning prediction model for urinary tract infection risks after a pediatric pyeloplasty operation, and a related system and equipment, and belongs to the technical field of urinary tract infection risk assessment after the pediatric pyeloplasty operation. According to the scheme of the application, cases of the pelvis forming operation due to unilateral hydronephrosis are screened from historical cases, the cases are divided into a training set and a verification set, a training sample is input into a TabNet deep learning network, a model is trained in a regression mode, the verification set is input into a trained TabNet model, evaluation, identification and calibration are carried out through cross validation, and the accuracy of the pelvis forming operation is improved. By inputting the clinical features of a to-be-predicted case into the model, a risk prediction result of clinical urinary tract infection diagnosed after an operation can be output. According to the diagnosis system, the accuracy, the precision and the area effect under an AUC curve after diagnosis feature modeling by applying the TabNet are better than those of Logistic Regression, SVM, Random Forest, XGBoost and LightGBM, feature engineering treatment does not need to be performed in advance, and the diagnosis system has important application value in urinary tract infection risk diagnosis after the pelvis forming operation of children.

Description

Technical Field

[0001] The present application relates to a method for constructing a machine deep learning prediction model for the risk of urinary tract infection after pyeloplasty in children and related systems and equipment, belonging to the technical field of risk assessment of urinary tract infection after pyeloplasty. Background Art

[0002] Ureteropelvic junction obstruction (UPJO) refers to an obstruction at the junction between the kidney and the ureter, resulting in reduced urine flow from the renal pelvis to the ureter. UPJO is the most common cause of pathological hydronephrosis in children, with an incidence of approximately 1:1500. Approximately 20% to 50% of children with UPJO eventually require surgical intervention. If not treated promptly, it may lead to hydronephrosis, chronic infection or urolithiasis, and often progressive renal insufficiency. Although the success rate of UPJO surgery is high, there are also many complications. The main complications of UPJO surgery are urinary tract infection (UTI), urine extravasation and leakage, pyelonephritis, bleeding, and recurrent UPJO.

[0003] Several studies have shown that UTI increases the risk of anastomotic restenosis after pyeloplasty in children with UPJO, and is considered to be one of the important factors for postoperative anastomotic restenosis, and the incidence of UTI remains high. UTI is mainly caused by Escherichia coli and is one of the most common types of infection in children. Previous studies have described risk factors for UTI in children, including male sex, weight, BUN level, recurrence of UTI within 3 months, long-term indwelling urinary catheter, double J tube, bilateral double J tube, etc. However, the development of predictive models for UTI after unilateral pyeloplasty is very limited, especially in large sample cohort studies. There is currently no artificial intelligence model for UTI after pyeloplasty.

[0004] Over the past decade, the rate of development of artificial intelligence has increased exponentially. The increase in machine learning processor power and the amount of available data has provided the foundation for the development of deep learning and artificial intelligence tools. The healthcare industry collects a large amount of data, such as medical records and patient test results. With the recent advances in artificial intelligence, these technologies are now increasingly used for clinical decision making. With the development of efficient integrated learning models, machine learning (ML) has made significant breakthroughs in the accuracy and efficiency of processing tabular data, and clinical medicine applications are particularly suitable for these methods.

[0005] For example, the ML algorithm can be used to analyze the prediction and diagnostic evaluation of bladder cancer, as well as the staging and recurrence prediction of renal cancer and testicular cancer. In terms of kidney disease, ML has also shown good predictive application effects. Although the application of artificial intelligence technology in the field of urology has gradually increased, there are still few reports on its application in the field of functional urology. A review article on the application of machine learning in healthcare published in the Lancet magazine in 2024 showed that medical staff and patients believed that ML risk prediction models could add benefits in the field of healthcare. But it is undeniable that the research and application of artificial intelligence technologies represented by ML may change the clinical practice of urology and improve the overall outcomes of patients.

[0006] TabNet is a deep learning (DL) end-to-end architecture with an encoder. It establishes a continuous multi-step architecture, which helps to select relevant features for decision making. In recent years, TabNet has also been increasingly used in clinical predictions, such as Zhang et al. used TabNet to predict the dose of tacrolimus after renal transplantation, Yu et al. developed a prediction model for rapatinib in breast cancer patients, Cahan et al. established a 30-day all-cause mortality model for patients with pulmonary embolism, Asadi-Pooya et al. used TabNet to distinguish different types of epilepsy, and Chen et al. developed an X-ray diagnosis model for breast cancer. However, when we searched for medical application literature on TabNet, it was still significantly less than the reports on traditional machine learning models. This may be related to the fact that most medical data structures are composed of small sample data sets from a single center, while deep learning models require large data samples for training and parameter tuning.

[0007] Although ML and DL have their own advantages and disadvantages in the development and application of medical modeling, there is currently no artificial intelligence model proposed for UTI after hydronephrosis surgery in children. This may result in our inability to provide early prediction and warning for patients after hydronephrosis surgery, and the occurrence of UTI cannot be treated promptly and effectively, increasing the risk of long-term functional effects.

[0008] Ureteropelvic junction stenosis is a common disease in children. Laparoscopic or open pyeloplasty is the standard surgical method for treating this disease. According to the experience of our center, the infection rate of more than 700 children who underwent unilateral pyeloplasty in the past 10 years exceeded 30% within six months after surgery, which has a great negative impact on children's postoperative recovery, quality of life and family economic burden. The current method of predicting postoperative infection relies heavily on empirical judgment and limited clinical parameters, and lacks a multifactorial clinical prognosis prediction model. Therefore, it is crucial to establish an accurate prediction model for urinary tract infection after pyeloplasty in children. Summary of the invention

[0009] In order to solve the above problems, a method for constructing a machine deep learning prediction model for the risk of urinary tract infection after pyeloplasty in children and related systems and equipment are provided. The accuracy, precision and area under the AUC curve of diagnostic feature modeling using TabNet are better than those of Logistic Regression, SVM, Random Forest, XGBoost and LightGBM, and no feature engineering processing is required in advance. It has important application value in the diagnosis of the risk of urinary tract infection after pyeloplasty in children.

[0010] According to one aspect of the present application, a method for constructing a machine deep learning prediction model for the risk of urinary tract infection after pyeloplasty in children is provided, the method comprising the following steps:

[0011] S1. Screen cases of patients who underwent pyeloplasty due to unilateral hydronephrosis from historical cases and divide the cases into training set and validation set;

[0012] S2, input the training samples into the TabNet deep learning network and train the model in regression mode;

[0013] S3. Input the validation set into the trained TabNet model and evaluate the identification through cross-validation, wherein the evaluation of identification includes accuracy, precision and recall, so as to obtain the machine deep learning prediction model of urinary tract infection risk after pyeloplasty in children.

[0014] Optionally, the case includes the following clinical features: 1) operation time; 2) post-operative drainage method: nephrostomy or DJ tube; 3) number of operations: first time or non-first time; 4) history of abdominal pain: yes or no; 5) causes of ureteropelvic junction stenosis: congenital stenosis, vagal vascular compression, ureteral polyps, ureteral valves, high ureters or ureteral rupture; 6) side of hydronephrosis: left or right; 7) age; 8) height; 9) BMI; 10) total protein Tp; 11) albumin Alb; 12) globulin Glb; 13) blood uric acid Ua; 14) blood creatinine Cr; 15) blood urea nitrogen Bun; 16) glomerular filtration rate eGFR; 17) neutrophil ratio N%; 18) lymphocyte ratio L%; 19) anterior-posterior diameter of renal pelvis APD.

[0015] It should be noted that according to the importance analysis of clinical features in the TabNet deep learning network in this application, it can be seen that gender, surgical method, weight, history of urinary tract infection, history of nephrostomy and renal parenchyma thickness are basically not important in the machine deep learning prediction model of urinary tract infection risk after nephroplasty in children in this application. Therefore, the clinical features can be simplified and the above 19 clinical feature indicators can be used. However, it should be noted that those skilled in the art can also add other clinical feature indicators.

[0016] In addition, it should be pointed out that the clinical characteristic information in the case is used for TabNet deep learning network training, and can also be used to input the clinical characteristic information of the case to be predicted into the trained machine deep learning prediction model of urinary tract infection risk after pyeloplasty in children, thereby outputting the risk prediction results of clinical urinary tract infection diagnosed in children after pyeloplasty. It can be understood that the case can also include other clinical characteristic information in addition to the above-mentioned records.

[0017] Optionally, the case includes the following clinical characteristics: 1) gender; 2) surgical method: laparoscopic or open; 3) operation time; 4) postoperative drainage method: pyelostomy or DJ tube; 5) number of operations: first time or non-first time; 6) history of urinary tract infection: yes or no; 7) history of pyelostomy: yes or no; 8) history of abdominal pain: yes or no; 9) causes of ureteropelvic junction stenosis: congenital stenosis, vagal vascular compression, ureteral polyps, ureteral valves, high ureters or Ureteral rupture; 10) Side of hydronephrosis: left or right; 11) Age; 12) Weight; 13) Height; 14) BMI; 15) Total protein Tp; 16) Albumin Alb; 17) Globulin Glb; 18) Blood uric acid Ua; 19) Blood creatinine Cr; 20) Blood urea nitrogen Bun; 21) Glomerular filtration rate eGFR; 22) Neutrophil ratio N%; 23) Lymphocyte ratio L%; 24) Anterior-posterior diameter of renal pelvis APD; 25) Renal parenchymal thickness.

[0018] Optionally, in the TabNet deep learning network in step S2: the data flows through the batch normalization BN layer and then enters the feature converter, the feature converter comprises n gated linear unit GLU blocks, each of the GLU blocks consists of a fully connected FC layer, a BN layer and a GLU activation module.

[0019] Optionally, the feature converter includes four gated linear unit (GLU) blocks, two of which are shared and the other two are independently set.

[0020] Optionally, the feature converter processes the batch normalized features and then outputs them to the attention converter through the segmentation layer in step i;

[0021] The attention converter includes a fully connected FC layer, a BN layer, a prior scale layer, and a sparsemax layer. The attention converter receives input from the segmentation layer, which is then processed by the fully connected FC layer and the BN layer, and then passed through the prior scale layer. The prior scale layer summarizes the feature magnitudes in the previous decision steps, expressed as:

[0022] Where γ represents the relaxation parameter;

[0023] The attention converter calculates a learnable mask vector M[i] based on the feature a[i-1] processed in the previous step for sparse feature selection:

[0024] M[i] = sparsemax(P[i-1]·h i (a[i-1]))

[0025] Among them, P[i-1] is the prior scale term, h i () is a trainable function that encapsulates the fully connected FC layer and BN layer. The sparsemax layer normalizes the coefficients to enforce sparse feature selection;

[0026] In order to further regulate the feature sparsity, the sparsity regularization term L in the form of entropy is adopted sparse :

[0027]

[0028] The mask vector is fed into the feature transformer, which processes the filtered features and separates them into two outputs:

[0029] [d[i],a[i]]=f i (M[i]·f)

[0030] Where d[i] represents the output of the decision step and a[i] provides information for subsequent steps.

[0031] Optionally, the screening of historical cases in step S1 must simultaneously meet the following requirements: 1) unilateral hydronephrosis undergoing pyeloplasty; 2) meeting the diagnosis of ureteropelvic junction stenosis; 3) complete clinical feature data;

[0032] The exclusion criteria for cases were any of the following: 1) bilateral hydronephrosis; 2) no conventional anti-infection regimen; 3) full ureteral dilatation; 4) duplicated kidney; 5) combined vesicoureteral reflux; 6) DRF of the affected kidney before surgery < 10%; 7) pediatric bladder and rectal dysfunction (BBD); 8) incomplete clinical feature data.

[0033] Optionally, the risk prediction result of clinical urinary tract infection diagnosed in the child after pyeloplasty is clinical urinary tract infection occurring within 6 months after the operation.

[0034] Optionally, in step S3, the identification is evaluated by 5-fold cross validation, wherein:

[0035]

[0036] TP, TN, FP, and FN stand for true positive, true negative, false positive, and false negative, respectively; P and N stand for positive and negative, respectively.

[0037] According to one aspect of the present application, a system for diagnosing the risk of urinary tract infection after pyeloplasty in children based on machine deep learning is provided, which is implemented based on any of the above-mentioned methods for constructing a machine deep learning prediction model for the risk of urinary tract infection after pyeloplasty in children, and the system comprises:

[0038] A data input module, wherein the data input module is used to input clinical characteristic data of the case;

[0039] A data processing module, which is used to execute the constructed machine deep learning prediction model;

[0040] A data output module is used to output the diagnosis results of urinary tract infection risk after pyeloplasty in children.

[0041] According to one aspect of the present application, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of any one of the above-mentioned methods for constructing a machine deep learning prediction model for the risk of urinary tract infection after pyeloplasty in children are implemented.

[0042] According to one aspect of the present application, a readable storage medium is provided, wherein the readable storage medium stores a computer program, wherein the computer program is suitable for being loaded by a processor to execute any of the steps of the above-mentioned method for constructing a machine deep learning prediction model for the risk of urinary tract infection after pyeloplasty in children.

[0043] The beneficial effects of this application include but are not limited to:

[0044] 1. According to the machine deep learning-based urinary tract infection risk diagnosis system for children after pyeloplasty in this application, the clinical prediction model of UTI after pyeloplasty provided in this application is helpful for early prediction and timely treatment, and ultimately reduces the risk of UTI affecting long-term function.

[0045] 2. According to the machine deep learning-based diagnosis system for urinary tract infection risk after pyeloplasty in children of the present application, the accuracy, precision and area under the AUC curve after modeling the diagnostic features using TabNet are better than Logistic Regression, SVM, Random Forest, XGBoost and LightGBM, and no feature engineering processing is required in advance. It has important application value in the diagnosis of urinary tract infection risk after pyeloplasty in children.

[0046] 3. According to the machine deep learning-based urinary tract infection risk diagnosis system for children after pyeloplasty in this application, the system provided by the present application scheme can predict the risk of urinary tract infection after pyeloplasty in children, provide prospective clinical intervention, reduce the incidence and re-obstruction rate of postoperative infection, improve the prognosis of pyeloplasty, and reduce health and economic burdens. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0048] Figure 1 This is a performance result diagram of the TabNet prediction model involved in Example 2 of the present application;

[0049] Figure 2 The result graph of the area under the ROC curve of different models involved in Example 2 of the present application;

[0050] Figure 3 This is a performance result diagram of the LightGBM model (25 features) involved in Example 2 of the present application;

[0051] Figure 4 This is a performance result diagram of the LightGBM model (18 features) involved in Example 2 of the present application;

[0052] Figure 5 This is a performance result diagram of the TabNet model (25 features) involved in Example 2 of the present application;

[0053] Figure 6 This is a graph of the LightGBM feature importance analysis results involved in Example 3 of the present application;

[0054] Figure 7 This is a diagram of the TabNet feature importance analysis results involved in Example 3 of the present application;

[0055] Figure 8 This is a schematic diagram of the TabNet encoder structure involved in Example 2 of the present application;

[0056] Fig. 9 This is a schematic diagram of the TabNet decoder structure involved in Example 2 of the present application;

[0057] Fig.10 This is a schematic diagram of the structure of the TabNet feature converter involved in Example 2 of the present application;

[0058] Fig.11 This is a schematic diagram of the structure of the TabNet attention converter involved in Example 2 of the present application. DETAILED DESCRIPTION

[0059] The present application is described in detail below in conjunction with examples, but the present application is not limited to these examples. Unless otherwise specified, the raw materials and catalysts in the examples of the present application are purchased through commercial channels.

[0060] In this application, data from more than 1,000 pediatric patients who underwent open or classic three-port laparoscopic unilateral pyeloplasty in the Urology Department of the Children's Hospital Affiliated to Capital Institute of Pediatrics from January 2012 to January 2023 were collected. After inclusion and exclusion, 764 patients were enrolled, and 25 clinical characteristics including general condition, medical history, surgical conditions, hydronephrosis measurement values, etiology, postoperative drainage methods, immunological indicators, renal function indicators and blood cell indicators were collected. After LightGBM was used to complete feature engineering, five traditional machine learning algorithms, including Logistic regression, SVM, random forest, XGBoosting and LightGBM, were used for modeling. At the same time, the deep learning TabNet algorithm was also used for feature importance analysis and modeling. In this application, the performance and feature analysis of different machine learning and deep learning models in the risk prediction of acute upper urinary tract infection (acute pyelonephritis) were compared, and a TabNet prediction model with very good prediction effect was constructed.

[0061] Among the 764 enrolled cases, 265 cases developed urinary tract infection after surgery, and 217 cases occurred within 3 months after surgery. Among the 265 UTI cases, 73 cases were diagnosed as acute upper urinary tract infection (acute pyelonephritis) and 192 cases were acute lower urinary tract infection. In the whole cohort, 41 cases (5.4%) had re-obstruction in the mid- and long-term follow-up after surgery, of which 22 cases had urinary tract infection in the early stage after surgery (within 3 months), while only 19 cases (3.8%) in the non-infected case group had long-term re-obstruction, and the difference was statistically significant (χ 2 =155.11, P<0.01). This shows that postoperative UTI has a significant impact on the occurrence of re-obstruction, indicating that the technical issues studied in this application scheme have important practical significance, and the prediction model provided by this application scheme is also of great value in controlling the risk of postoperative UTI in actual pediatric patients.

[0062] In this application, the first machine learning prediction model for urinary tract infection after pyeloplasty in children was proposed, the important role of eGFR and ALB in the prediction of UTI after pyeloplasty in children was proposed, the role of feature engineering in model building was explored, and the modeling performance of traditional machine learning LightGBM and deep learning TabNet algorithms were compared.

[0063] The prediction model provided by this application can be used to predict the risk of urinary tract infection after pyeloplasty in children, provide prospective clinical intervention, reduce the incidence of postoperative infection and re-obstruction rate, improve the prognosis of pyeloplasty, and reduce health and economic burden. Given its strong potential, it is worth further research and implementation in clinical practice to improve patient prognosis.

[0064] This study was reviewed and approved by the Institutional Review Board of the Capital Institute of Pediatrics. This study was reviewed by the Ethics Committee of this institution, and the ethics approval number is SHERLLM2023041. During the data acquisition period (June 1, 2023 to December 31, 2023), a total of 1033 cases of pediatric hydronephrosis were obtained. After applying the inclusion and exclusion criteria, a total of 764 cases were included in the cohort. Case inclusion criteria: (1) unilateral hydronephrosis undergoing pyeloplasty (2) in accordance with the diagnosis of ureteropelvic junction stenosis in accordance with the EAU and the "Expert Consensus on the Diagnosis and Treatment of Congenital Hydronephrosis" (3) complete clinical data and cooperation with postoperative follow-up. Case exclusion criteria: (1) bilateral hydronephrosis (2) no conventional anti-infection regimen (3) full ureteral dilatation (4) duplicate kidney (5) combined vesicoureteral reflux (6) preoperative dilated renal function (DRF) of the affected kidney <10% (7) children with bladder and rectal dysfunction (BBD) (8) incomplete clinical data. Patients were divided into 2 cohorts in an 8:2 ratio: 1 training set (n=612) and 1 validation set (n=152).

[0065] The data collected for the included cases included the general condition of the children (sex, age, height and weight, body mass index (BMI), conditions before pyeloplasty (history of abdominal pain, history of nephrostomy, history of urinary tract infection), surgical conditions (number of surgeries (first / non-first), operation time, surgical method (laparoscopic / open), side of hydronephrosis), anteroposterior diameter of the renal pelvis and renal parenchyma thickness measured by Doppler color ultrasound before surgery, causes of ureteropelvic junction stenosis (congenital stenosis, vagal vascular compression, ureteral polyps, ureteral valves, high ureters, ureteral rupture), drainage method after surgery (nephrostomy / DJ tube), immunological indicators (total protein, albumin, globulin), renal function indicators (serum creatinine, blood urea nitrogen, glomerular filtration rate, blood uric acid) and blood cell indicators (neutrophil ratio, lymphocyte ratio).

[0066] The outcome label of this study was clinical urinary tract infection (UTI) diagnosed after pyeloplasty, and the time of occurrence was defined as UTI occurring within 6 months after surgery. The diagnosis of UTI was based on: ① 2015 European Association of Urology (EAU) guidelines for urinary tract infection; ② 2016 evidence-based guidelines for the diagnosis and treatment of urinary tract infection by the Nephrology Group of the Pediatric Branch of the Chinese Medical Association.

[0067] Cefuroxime sodium (20 mg / kg) was intravenously infused once within 0.5 hours before surgery to prevent infection; cefuroxime sodium was intravenously infused for 5 to 7 days after surgery, with a daily dose of 30 to 100 mg / kg, divided into 3 to 4 doses. The next treatment plan was adjusted according to clinical symptoms and urine routine monitoring. Prophylactic anti-infection plan during DJ tube indwelling period: Since there are currently no relevant guidelines and consensus on prophylactic anti-infection for DJ tube indwelling in children, our center adopts the EAU and AUA guidelines on VUR anti-infection and the recommended plan of the consensus of experts from the Chinese Society of Pediatric Urology: ① Infants under 2 months old: Cefaclor suspension, taken orally once a day before bedtime, the dose is 1 / 3 of the therapeutic dose. ② 2 months to 7 years old: Co-trimoxazole (complex sulfamethoxazole), the dose is 1 / 3 of the therapeutic dose. ③ Over 7 years old: Amoxicillin clavulanate potassium: taken orally once a day before bedtime, the dose is 1 / 3 of the therapeutic dose. During the period of DJ tube indwelling, urine routine examination and urine smear test should be reviewed every 2 weeks, anti-infection regimen should be adjusted in time, and drug resistance monitoring should be carried out.

[0068] This study used 6 machine learning and deep learning algorithms to establish a prediction model for urinary tract infection after pyeloplasty and compared the model performance of different algorithms. This study used 5 common machine learning algorithm models including Logistic regression, support vector machine SVM, random forest Random-Forest, extreme gradient boosting tree model XGBoosting and lightweight gradient boosting tree LightGBM. Accuracy, precision, recall and area under the receiver operating characteristic (ROC) curve (AUC) were used to evaluate the prediction accuracy and conformity of the model. Discrimination and calibration were evaluated by 5-fold cross validation. The calculation formulas for accuracy, precision and recall are as follows:

[0069]

[0070] TP, TN, FP, and FN stand for true positive, true negative, false positive, and false negative, respectively; P and N stand for positive and negative, respectively.

[0071] The present application scheme is described below through specific embodiments.

[0072] Example 1 Dataset Structure

[0073] The research data in this application were derived from pediatric patients who underwent surgery for hydronephrosis in a single center within 10 years. According to the exclusion criteria, the data of 764 patients were finally included for retrospective cohort study. Among the 764 enrolled cases, 265 developed urinary tract infection after surgery, of which 217 occurred within 3 months after surgery and 48 occurred between 3 and 6 months after surgery. Among the 265 cases of urinary tract infection, 73 were diagnosed with acute upper urinary tract infection (acute pyelonephritis) and 192 were acute lower urinary tract infection. Among the upper urinary tract infection cases, 32 were diagnosed with acute focal bacterial nephritis and 7 were diagnosed with renal abscess. New renal scarring was found by DMSA examination.

[0074] Each case in this cohort was followed up for a long time (1 to 8.5 years, median 5.5 years), and the incidence of urinary tract infection and re-obstruction after surgery was collected and recorded. The results showed that 41 cases (5.4%) in the whole cohort had re-obstruction in the mid- to long-term follow-up after surgery, of which 22 cases had urinary tract infection in the early stage after surgery (within 3 months), while nearly 19 cases (3.8%) in the group without infection after surgery had re-obstruction in the long term, and the difference was statistically significant (χ 2 =155.11, P<0.01). This shows that early postoperative urinary tract infection has a significant impact on the occurrence of ureteropelvic junction re-obstruction in the medium and long term.

[0075] Cases were included in this study, and 25 clinical features were collected, including demographic characteristics, biochemical renal function test indicators, immunological indicators, and special hydronephrosis indicators. The clinical baseline and statistical description of the cohort are shown in Table 1. The training cohort and the test cohort were randomly divided into 4:1 groups, and there was no statistical difference in the distribution of the 25 features, as shown in Table 2. Enumeration data were expressed as "cases", and the inter-group comparisons were performed using the χ2 test or Fisher's exact probability method. Quantifiable data were expressed as M[Q1,Q3], and the inter-group comparisons were performed using the independent sample t test or rank sum test. P value <0.05 indicated statistical significance.

[0076] The clinical characteristics corresponding to each row in Tables 1 and 2 are as follows: 1) gender; 2) surgical method: laparoscopic or open; 3) operation time; 4) postoperative drainage method: pyelostomy or DJ tube; 5) number of operations: first time or non-first time; 6) history of urinary tract infection: yes or no; 7) history of pyelostomy: yes or no; 8) history of abdominal pain: yes or no; 9) causes of ureteropelvic junction stenosis: congenital ureteral stricture, vagalvein compression, ureteral polyps, ureteral valves, high ureter or ureteral rupture rupture); 10) side of hydronephrosis: left L or right R; 11) age; 12) weight; 13) height; 14) BMI; 15) total protein Tp; 16) albumin Alb; 17) globulin Glb; 18) blood uric acid Ua; 19) blood creatinine Cr; 20) blood urea nitrogen Bun; 21) glomerular filtration rate eGFR; 22) neutrophil ratio N%; 23) lymphocyte ratio L%; 24) anterior-posterior diameter of renal pelvis APD; 25) renal parenchymal thickness.

[0077] Table 1 Baseline characteristics of infection and non-infection groups

[0078]

[0079] Table 1 continued

[0080]

[0081] Table 2 Univariate comparison of training cohort and validation cohort

[0082]

[0083] Table 2 continued

[0084]

[0085] Example 2 TabNet prediction model

[0086] TabNet is a deep neural network tailored for tabular data processing. Its encoder architecture (e.g. Figure 8As shown in Figure 1, the TabNet encoder is composed of a feature transformer, an attentive transformer, and a feature mask. The split block splits the processed output representation for the attentive transformer in the subsequent steps and the overall use of the output. For each step, the feature selection mask provides interpretable information about the model's functionality, and the masks can be aggregated to obtain global feature important properties. In each step i, the processed information from the previous step i-1 guides feature selection and generates aggregated feature representations for overall decision making. The model inputs a data batch with a specified batch size B and D-dimensional features without global feature selection or normalization, where B is the batch size and D is the feature dimension. The BN operation, i.e. BatchNormal normalization, is performed to obtain BN_Features with a dimension of B×D. Then, based on serialized multi-step processing, there are a total of N_steps rounds of decision making, and BN_Features is input in each round. The features selected in the i-th round are determined by the information passed from the i-1th round. The selected features are then feature processed and finally the results of all rounds of decision feature processing are fused, and the corresponding activation function outputs are connected according to the task type.

[0087] The data flows through the batch normalization layer (BN layer) and enters the feature transformer. The transformer contains n gated linear unit (GLU) blocks, each of which consists of a fully connected (FC) layer, a BN layer, and a GLU activation. As a specific implementation, Fig.10 As shown, when four GLU blocks are used, two are shared and two are independent, resulting in a total of 4 layers of the network, of which 2 layers are shared by all decision steps and 2 layers are related to the decision steps, thereby optimizing parameter efficiency and robustness, and adding skip connections between consecutive blocks to maintain information flow stability.

[0088] The feature converter processes the batch normalized features and outputs them to the attention converter through the split layer (Split) at step i. Fig.11 As shown in the figure, the Attentive transformer integrates a series of layers: FC, BN, Prior scales, and sparsemax. First, it receives input from the split layer, processes it through the FC layer and the BN layer, and then passes through the Prior scale layer. The Prior scale layer summarizes the feature magnitudes in the previous decision step, expressed as:

[0089]

[0090] Where γ represents the relaxation parameter.

[0091] like Fig.11 As shown, the attention transformer calculates a learnable mask vector M[i] based on the feature a[i-1] processed in the previous step for sparse feature selection:

[0092] M[i] = sparsemax(P[i-1]·h i (a[i-1]))

[0093] Where P[i-1] is the prior scale term and hi() is a trainable function that encapsulates the FC layer and the BN layer. The sparsemax layer normalizes the coefficients to enforce sparse feature selection. In the Attentive Transformer, a single layer map is modulated using the previous scale information that summarizes the usage of each feature before the current decision step, and sparsemax is used to normalize the coefficients to achieve sparse selection of significant features.

[0094] In order to further regulate the feature sparsity, a sparsity regularization term (L sparse ):

[0095]

[0096] Where ∈ is a small constant to ensure the sparsity of the regularization term, and no specific requirements are made in the model. In the model of the present application, ∈ is set to e^-20.

[0097] The mask vector is fed into the feature transformer, which processes the filtered features and separates them into two outputs:

[0098] [d[i],a[i]]=f i (M[i]·f)

[0099] Where d[i] represents the output of the decision step and a[i] provides information for subsequent steps.

[0100] TabNet provides interpretability through local and global feature explanations. The model aggregates the step outputs to produce a scalar that represents the importance of the step to the final result:

[0101]

[0102] The local feature importance comes from summing the step contributions of a single sample b. The global aggregate feature importance mask Magg (b,j)is the normalized feature contribution, calculated as follows:

[0103]

[0104] like Fig. 9 The following is a schematic diagram of the TabNet decoder, where the input is the encoded representation of the data (Encoded representation) encoded by the encoder, and the output is the reconstructed features (Reconstructed features), and each step consists of a feature transformer (Feature transformer). TabNet eliminates the need for feature engineering and is an improvement over traditional methods such as random forests and LightGBM. Unlike previous methods, TabNet can directly process tabular data without manually selecting or converting features. This greatly reduces the workload of preprocessing and can effectively process unprocessed raw data.

[0105] Example 2: Comparison between feature engineering and traditional machine learning modeling

[0106] 1) Feature Engineering

[0107] Traditional machine learning requires feature engineering before modeling. Feature engineering first sorts the importance of all features according to the basic model, performs feature screening, sets up cross-validation, and uses network parameter search methods, usually using accuracy as the criterion, to include features that have a significant impact on model results into the final model. However, the deep learning model (TabNet) does not require feature screening for feature engineering. It can input all features into the final model and only sort and analyze the importance of all features.

[0108] The results show that feature engineering can significantly improve the predictive performance of traditional machine learning models. Taking LightGBM as an example, the test results of the LightGBM model before feature engineering showed an accuracy of 0.70, a precision of 0.69, a recall of 0.70, and an area under the ROC curve of 0.74. The test results of the LightGBM model after feature engineering showed an accuracy of 0.71, a precision of 0.71, a recall of 0.71, and an area under the ROC curve of 0.78.

[0109] 2) Prediction results of different models

[0110] The performance comparison results of the prediction models constructed using different algorithms are shown in Table 3. Figure 1 As shown, the ROC results of different models are as follows Figure 2 shown.

[0111] Table 3 Performance comparison of prediction models constructed using different algorithms

[0112]

[0113] The performance ranking of six machine learning and deep learning algorithms shows that the order of model performance from low to high is: Logistic Regression, SVM, Random Forest, XGBoost, LightGBM without feature engineering, LightGBM with feature engineering, and finally TabNet.

[0114] The performance of the LightGBM model (25 features) before feature engineering shows Accuracy: 0.6959, Precision: 0.6933, Recall: 0.6959, ROC AUC: 0.7444. The results are as follows Figure 3 The performance of the LightGBM model (18 features) after feature engineering on the test set shows Accuracy: 0.7059, Precision: 0.7043, Recall: 0.7059, ROC AUC: 0.7794. The results are as follows Figure 4 The TabNet model (25 features) test set performance shows Accuracy: 0.7516, Precision: 0.8571, Recall: 0.4762, ROC AUC: 0.7918. The results are as follows Figure 5 shown.

[0115] By comparison, it can be seen that the prediction model using TabNet in this application has the best effect.

[0116] Example 3 Feature importance analysis and visualization results of LightGBM and TabNet

[0117] 1) LightGBM feature importance analysis

[0118] The results are as follows Figure 6 As shown. The feature importance results based on the LightGBM model show that the estimated glomerular filtration rate (eGFR) is the most influential feature, followed by blood urea nitrogen value (Bun), serum albumin (ALB), neutrophil ratio (n), operation time (time), serum creatinine value (cr), renal parenchymal thickness (Renal), serum globulin (GLB), blood uric acid value (ua) and lymphocyte ratio (l); in addition, the anteroposterior diameter of the renal pelvis (apd), body mass index (bmi), serum total protein (tp), age (age), weight (weight), drainage method (drainge), height (height), and surgical method (surgery) also have a certain influence in the prediction model.

[0119] 2) TabNet feature importance analysis

[0120] The results are as follows Figure 7 As shown. The feature importance results based on the Tabnet model showed that serum albumin (ALB) was the most influential feature, followed by estimated glomerular filtration rate (eGFR), serum globulin (GLB), serum total protein (tp), serum creatinine (cr), lymphocyte ratio (l), anteroposterior diameter of renal pelvis (apd), drainage method (drainge), surgical side (side), and blood urea nitrogen value (Bun); in addition, age (age), body mass index (bmi), height (height), history of abdominal pain (abdominalgia), blood uric acid (ua), cause (cause), number of operations (number of oerations), and neutrophil ratio (n) also had a certain influence in the prediction model.

[0121] In summary, LightGBM, or lightweight gradient boosting machine, is an efficient tree-based learning algorithm that optimizes computing speed and memory usage by building histograms and adopting a leaf-first strategy, significantly improving the efficiency of processing large-scale data. In addition, LightGBM supports efficient parallel computing, which enables it to perform well in big data environments. However, this model may not be as effective as a neural network model on data sets with higher feature dimensions or more categorical features because it performs less automatic feature crossover and complex pattern recognition.

[0122] TabNet, through its unique encoder and sparse learning mask mechanism, can effectively perform feature selection and representation learning on tabular data. This mechanism enables TabNet to capture complex data relationships without the need for intensive feature engineering, especially in terms of feature interaction and nonlinear pattern recognition. In addition, TabNet's online learning capability provides the model with the possibility of continuous updating and adapting to dynamically changing data environments.

[0123] The feature importance analysis of the LightGBM model and the Tabnet model showed that eGFR and alb were common important features of the two models. The LightGBM model screened and ranked the 25 features and the visualization analysis results using the SHAP value showed that eGFR, blood urea nitrogen (Bun), and serum albumin concentration (Alb) were the top three important features. The feature importance analysis results of the Tabnet model showed that serum albumin concentration (Alb), eGFR, and Glb were the top three important features. The two indicators, eGFR and ALB, showed high importance in both the LightGBM and TabNet models.

[0124] Estimated Renal glomerular filtration rate (eGFR) can be assessed by measuring substance clearance, blood circulation markers, and formulas for estimating glomerular filtration rate. In this study, the eGFR of all cases was calculated using the updated Schwartz bedside eGFR formula (CKiDcr). Both the LightGBM and TabNet models showed that eGFR was significant and contributed greatly to the predictive ability of the model, which has important clinical significance for predicting infection after pyeloplasty. Previous studies have also confirmed that eGFR can be used as a predictor of urinary tract infection in patients with diabetes. Although none of the cases in this study had a history of systemic disease, considering that all children underwent urinary tract devascularization and reconstruction surgery, which is invasive and will damage the urethral mucosa, making it easy for bacteria to enter, thereby increasing the risk of urinary tract infection, eGFR does not necessarily need to reach a critical value to be clinically significant.

[0125] Our results show that the lower the eGFR, the higher the risk of urinary tract infection, and the lower eGFR level may be related to the severity of obstructive hydronephrosis. Previous studies have shown that the maximum anterior-posterior diameter of the renal pelvis >15mm, accompanied by dilatation of the surrounding calyces and changes in the renal parenchyma are high-risk factors for urinary tract infection. Urinary tract obstruction may also reduce the self-cleaning function of the urinary tract in children, and the renal system promotes bacterial growth and reproduction due to excessive residual urine. Studies have shown that urine retention caused by UPJO provides time and opportunity for bacteria to adhere to urothelial cells, proliferate and infect the host.

[0126] In the past, serum albumin levels were generally considered an indicator of nutritional status, but recent literature has shown that serum albumin concentrations, disease severity, and mortality are closely related. Hypoalbuminemia can be considered an independent risk factor for disease morbidity and mortality, especially for infectious diseases. Low serum albumin levels are associated with an increased risk of death in patients with severe sepsis. Low albumin concentrations reflect low serum protein levels, resulting in low serum bactericidal activity and susceptibility to bacterial infections. Falguera et al.'s study believes that low serum albumin is a predictive factor for patients with infections (such as community-acquired pneumonia). Since the ALB test results in this study were all obtained within 3 days before surgery, the numerical impact caused by surgery can be reduced. In recent years, some studies have also pointed out that albumin levels are also an indicator of inflammatory status. Lee et al.'s study showed that hypoalbuminemia may be a reflection of baseline nutritional levels or related to inflammatory changes during sepsis.

[0127] Although the exact mechanism has not been fully elucidated, serum albumin has protective effects, such as antioxidant activity, maintaining physiological homeostasis, and anti-inflammatory effects. Studies by Li et al. showed that children with acute glomerulonephritis (AGN) with low serum albumin levels were more susceptible to urinary tract infections. In recent years, some scholars have also proposed that different biological test results can be combined to predict the risk of urinary tract infections. For example, the serum procalcitonin / albumin ratio can be used for early diagnosis and prediction of urosepsis and febrile urinary tract infections. Among patients with urosepsis, patients with a higher procalcitonin / albumin ratio are more likely to develop urosepsis shock. The urea nitrogen / albumin ratio can be used to predict febrile UTI in infants, identify children at high risk of bacteremia, and facilitate management decisions for infants with febrile UTI. The results of this study show that ALB levels should be monitored in children undergoing pyeloplasty, early identification of children at high risk of postoperative urinary tract infections, and timely clinical management decisions should be made.

[0128] The TabNet model provided in this application can significantly improve the prediction performance of the model through 25 clinical characteristics of the case. It can provide early warning, optimize treatment, and improve prognosis for the risk of urinary tract infection in pediatric patients who have undergone pyeloplasty, and has important application scenarios.

[0129] The above is only the embodiment of the present application, and the protection scope of the present application is not limited by these specific embodiments, but is determined by the claims of the present application. For those skilled in the art, the present application can have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the technical ideas and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for constructing a machine deep learning prediction model for the risk of urinary tract infection after pyeloplasty in children, characterized in that: The method comprises the following steps: S1. Screen cases of patients who underwent pyeloplasty due to unilateral hydronephrosis from historical cases and divide the cases into training set and validation set; S2, input the training samples into the TabNet deep learning network and train the model in regression mode; S3. Input the validation set into the trained TabNet model and evaluate the identification through cross-validation, wherein the evaluation of identification includes accuracy, precision and recall, so as to obtain the machine deep learning prediction model of urinary tract infection risk after pyeloplasty in children.

2. The method for constructing a machine deep learning prediction model for the risk of urinary tract infection after pyeloplasty in children according to claim 1, characterized in that: The cases described included the following clinical features: 1) Operation time; 2) Postoperative drainage method: pyelostomy or DJ tube; 3) Number of surgeries: first or non-first; 4) History of abdominal pain: yes or no; 5) Causes of ureteropelvic junction stenosis: congenital stenosis, vagal vascular compression, ureteral polyps, ureteral valves, high ureter or ureteral rupture; 6) Side of hydronephrosis: left or right; 7) Age; 8) Height; 9) BMI; 10) Total protein Tp; 11) Albumin Alb; 12) Globulin Glb; 13) Serum uric acid Ua; 14) Serum creatinine Cr; 15) Blood urea nitrogen Bun; 16) Glomerular filtration rate eGFR; 17) Neutrophil ratio N%; 18) Lymphocyte ratio L%; 19) Anterior-posterior diameter of renal pelvis APD.

3. The method for constructing a machine deep learning prediction model for the risk of urinary tract infection after pyeloplasty in children according to claim 1, characterized in that: In the TabNet deep learning network in step S2: the data flows through the batch normalization BN layer and then enters the feature converter, the feature converter includes n gated linear unit GLU blocks, each of the GLU blocks consists of a fully connected FC layer, a BN layer and a GLU activation module; Optionally, the feature converter includes four gated linear unit (GLU) blocks, two of which are shared and the other two are independently set.

4. The method for constructing a machine deep learning prediction model for the risk of urinary tract infection after pyeloplasty in children according to claim 3, characterized in that: The feature converter processes the batch normalized features and then outputs them to the attention converter in step i through the segmentation layer; The attention converter includes a fully connected FC layer, a BN layer, a prior scale layer, and a sparsemax layer. The attention converter receives input from the segmentation layer, which is then processed by the fully connected FC layer and the BN layer, and then passed through the prior scale layer. The prior scale layer summarizes the feature magnitudes in the previous decision steps, expressed as: Where γ represents the relaxation parameter; The attention converter calculates a learnable mask vector M[i] based on the feature a[i-1] processed in the previous step for sparse feature selection: M[i]=sparsemax(P[i-1]·h i (a[i-1])) Among them, P[i-1] is the prior scale term, h i () is a trainable function that encapsulates the fully connected FC layer and BN layer. The sparsemax layer normalizes the coefficients to enforce sparse feature selection; In order to further regulate the feature sparsity, the sparsity regularization term L in the form of entropy is adopted sparse : The mask vector is fed into the feature transformer, which processes the filtered features and separates them into two outputs: [d[i],a[i]]=f i (M[i]·f) Where d[i] represents the output of the decision step and a[i] provides information for subsequent steps.

5. The method for constructing a machine deep learning prediction model for the risk of urinary tract infection after pyeloplasty in children according to claim 1, characterized in that: The screening of historical cases in step S1 must simultaneously meet the following requirements: 1) unilateral hydronephrosis undergoing pyeloplasty; 2) meeting the diagnosis of ureteropelvic junction stenosis; 3) complete clinical characteristics data; The exclusion criteria for cases were any of the following: 1) bilateral hydronephrosis; 2) no conventional anti-infection regimen; 3) full ureteral dilatation; 4) duplicated kidney; 5) combined vesicoureteral reflux; 6) DRF of the affected kidney before surgery < 10%; 7) pediatric bladder and rectal dysfunction (BBD); 8) incomplete clinical feature data.

6. The method for constructing a machine deep learning prediction model for the risk of urinary tract infection after pyeloplasty in children according to claim 1, characterized in that: The risk prediction result of clinical urinary tract infection diagnosed after pyeloplasty in children is clinical urinary tract infection occurring within 6 months after surgery.

7. The method for constructing a machine deep learning prediction model for the risk of urinary tract infection after pyeloplasty in children according to claim 1, characterized in that: In step S3, the identification is evaluated by 5-fold cross validation, where: Accuracy Accuracy Recall TP, TN, FP, and FN stand for true positive, true negative, false positive, and false negative, respectively; P and N stand for positive and negative, respectively.

8. A urinary tract infection risk diagnosis system for children after pyeloplasty based on machine deep learning, characterized in that: The method is based on the method for constructing a machine deep learning prediction model for the risk of urinary tract infection after pyeloplasty in children according to any one of claims 1 to 7, and the system comprises: A data input module, wherein the data input module is used to input clinical characteristic data of the case; A data processing module, which is used to execute the constructed machine deep learning prediction model; A data output module is used to output the diagnosis results of urinary tract infection risk after pyeloplasty in children.

9. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the method for constructing a machine deep learning prediction model for the risk of urinary tract infection after pyeloplasty in children are implemented as described in any one of claims 1 to 7.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, which is suitable for being loaded by a processor to execute the steps of the method for constructing a machine deep learning prediction model for the risk of urinary tract infection after pyeloplasty in children as described in any one of claims 1 to 7.

Images