Printer Friendly

Application of data mining techniques to explore predictors of upper urinary tract damage in patients with neurogenic bladder.


Patients with neurogenic bladder (NGB) not only suffer from the inconvenience caused by lower urinary tract symptoms, e.g., urinary incontinence and retention, but also have the risk of upper urinary tract damage (UUTD) including vesicoureteral reflux, ureteral dilation, and hydronephrosis (1). The principal aims of treatment to NGB patients are preservation of the upper urinary tract function, avoidance of the risk factors, prevention of upper urinary tract complications, and maximization of patients' life expectancy. Currently, considerable morbidity (20-80%) of UUTD in NGB patients has been reported in a variety of studies (2).

UUTD is one of the most severe complications of NGB; therefore, early diagnosis is very important for a timely neurological management to avoid any further damage. Generally, upper urinary tract damage is detected by cystourethrogram, video urodynamic studies, or other image examinations. However, in many undeveloped or developing areas, it is not easy for patients to get access to these examinations conveniently due to economic reasons or equipment shortage. Therefore, a more convenient and economical method to predict the risk of UUTD of NGB is needed.

Research demonstrated that urodynamic changes, bladder management and anatomical structure changes of bladder and ureter are related to upper urinary tract deterioration of NGB. However, the roles of these factors in detection of UUTD for NGB patients are still controversial, and the interaction of these factors have not been well considered. As a result, current methods are still not precise enough for the risk prediction of UUTD in NGB patients (3-7).

Decision tree is one of the most commonly used data mining techniques. It can establish a set of hierarchical rules from a group of disorderly and irregular cases. Based on these rules, a model can be established to predict the characteristics of a disease effectively with accessible and precise results. It has already been used in various medical fields such as the exploration of risk factors for type 2 diabetes, cancer, and high blood pressure (8-12).

In this research, a decision tree model for the early prediction of neurogenic UUTD was established by analyzing cases of NGB retrospectively.

Material and Methods


A total of 112 patients with NGB were recruited from the Third Hospital affiliated to Sun Yat-Sen University, Nanfang Hospital of Southern Medical University, and Foshan City First People's Hospital (Guangdong, China) between January 2006 and December 2015. This retrospective study was approved by the Nanfang Hospital's ethics committee. Inclusion criteria were that patients should have NGB or secondary neuropathic bladder due to a variety of reasons (ICD-10 code N31.901). Patients were excluded if: 1) they had never accepted imaging tests of the urinary tract; 2) renal insufficiency was not secondary to NGB; 3) had no regular visit to hospital; 4) had no regular urodynamic tests; 5) did not have complete records. Patients were divided into two groups, the case group (n=34) and the control group (n=78), depending on the development of UUTD during hospitalization. UUTD was diagnosed mainly based on the results of imaging examinations, which were found in previous medical records. The diagnosis criteria were: 1) renal insufficiency in diagnosis; 2) results of imaging examinations on urinary tract showed that renal pelvis and/or ureter were full-filled or dilated; and 3) results of imaging examinations on urinary tract showed vesicoureteral reflux (VUR).

Data collection

As a case-control retrospective study, medical records of patients were reviewed to summarize the history of urinary tract infection, bladder managements, conservative treatments and urodynamic findings. Patients who were diagnosed with urinary tract infection were considered as having history of urinary tract infection. Bladder management methods, including manual micturation by increasing abdominal pressure (like tapping or extruding abdomen), voluntary urination, indwelling urethral catheterization, clean intermittent catheterization and cystostomy, were used to empty bladder in voiding period. Conservative treatments with non-surgical and non-pharmacologic methods, including bladder training, pelvis floor muscle exercises, and pelvis floor electrostimulation, were used to promote rehabilitation of bladder function in voiding period.

All patients accepted the urodynamic tests with Delphis[TM] Urodynamic Analyzer (94-R01-BT, Canada). In the urodynamic tests, an 8F catheter was first inserted into the bladder through the urethra; then, the bladder was filled with normal saline at 20-30 mL/min in the beginning and gradually up to 50-60 mL/min. The urine samples were collected at the end of filling period. During the filling and voiding periods, patients' bladder sensation and reaction to the tests were recorded. In this study, detrusor activity, bladder sensation, bladder capacity, bladder compliance, urethral function, maximum bladder capacity, Pves max (maximum intravesical pressure), Pabd max (maximum abdominal pressure), and Pdet max (maximum detrusor pressure) were recorded in the filling phase. Detrusor contractility, urethral function, and Pdet max were recorded in the voiding phase during the urodynamic tests. In all cases, urodynamic test results acquired at the onset of UUTD and at the last hospital visit for the case and control groups, respectively, were recorded.

Statistical analysis

Data analyses were performed using SPSS software (version 22.0, IBM, USA). The t-test and chi-square test were used to assess differences for each indicator between the case and control group. Data are reported as the frequency, proportion, and means [+ or -] SD.

Classification and regression tree (CART) method of decision was applied to develop the model. UUTD was used as the dependent variable and history of urinary tract infections, bladder management, conservative treatment, and urodynamic findings were used as the independent variables. Cross validation was used to verify the data processing results. The regression tree model was built using the standard CRT algorithm in SPSS. The maximum tree depth and surrogates were automatically set by the software. The minimum number of cases to determine parent node and child node were 10 and 5, respectively. Gini impurity was used for impurity measurement and the minimum change in improvement was set as the default value. Tree pruning was not performed.

Sensitivity, specificity, positive predictive value, negative predictive value, accuracy and area under curve (AUC) were used to describe the predictive value of the decision tree model. The Kappa consistency analysis was applied to test the results. A P value less than 0.05 was considered statistically significant.


The study population consisted of 112 patients with NGB. Of those, 34 had spinal injury, 32 had diabetes, 20 had cerebrovascular accident, 6 had lumbar disc herniation, 4 had spinal pelvic surgery, 2 had craniocerebral injury, 2 had spinal stenosis, 1 had Green-Barre syndrome, 1 had multiple sclerosis, 1 had viral encephalitis, 1 had Parkinson disease, and 8 with unknown cause. The case and control groups were created according to presence or absence of UUTD, respectively (Table 1).

A decision tree model was generated to analyze several indicators for risk of UUTD. Urethra function, Pabd max, Pves max, and gender were selected as the screening indicators, and cut-off values for Pabd max and Pves max were 14 and 89, respectively (Figure 1). Extraction of the higher risk of neurogenic UUTD established in this study indicated two classification rules, one is normal urethra function, Pabd max > 14 cm [H.sub.2]O, and Pves max [less than or equal to] 89 cm [H.sub.2]O, and the other is normal urethra function, Pabd max > 14 cm [H.sub.2]O, Pves max > 89 cm [H.sub.2]O and female gender (Table 2).

The evaluation indices for the risk of UUTD using the decision tree are illustrated in Table 3. The sensitivity, specificity, and overall accuracy in predicting risk of UUTD were 85.3, 84.6, and 84.8%, respectively. In addition, results showed that the AUC was 0.901 (95% CI = 0.844-0.958) (Table 3, Figure 2).

The accuracy of prediction was measured by the cross-validation method, including risk assessment and its standard error. The risk assessment of this model was 23.2%, which is the percentage of cases that would be wrongly classified (Table 4). Kappa test was used to analyze the consistency of the predicted results. The Kappa value was 0.661 (P< 0.001), showing a good fitness between the prediction results and the actual morbidity of UUTD.


UUTD in NGB includes a series of complications such as pyeloureterectasis, hydronephrosis, VUR and renal insufficiency (13). It is demonstrated that the anatomic structure and physiological function of intramural ureter, bladder triangle and ureter undertake the responsibility for anti-reflux. Urodynamic changes, bladder management model, complications of urinary system and some other factors may impair the anti-reflux mechanism and induce the urine in bladder to flow back into ureter. The refluxed urine increases the pressure of renal pelvis and calyces, expanding them, and finally leading to chronic renal failure (14). Until now, most research focused on the influence of a single factor on the UUTD in NGB; few studies are on the risk prediction of neurogenic UUTD.

Decision tree technique, which could establish a set of classification rules from a group of disorder and irregular cases, is a commonly used data mining method in clinical research. Technically, a tree diagram is used to show all the possible outcomes and classify the study objects accurately. Recently, it has been used to excavate major risk factors in some diseases (8-12). The present study established a decision tree model to excavate the major risk factors for UUTD in NGB patients. The results indicated that urethra function in the filling phase was a major predictive variable; also, Pabd max, Pves max, and gender were associated with the occurrence of the UUTD in NGB patients.

Under normal urethra function, urethral closure pressure can still maintain a positive value, which would induce urinary continence during filling phase even with increased abdominal pressure. However, if urethra function is incompetent, urine leaks out and increases abdominal pressure even if there is no detrusor contractility. In patients, the injured micturition reflex nerve would lead to the inactive sphincter contractility during bladder filling phase or the abnormal urethral closure pressure, both of which would lead to an incompetent urethra function. The leakage of urine from urethra could decrease Pves in filling phase and reduce the reflux of urine from bladder to ureter. Further, the incompetent urethra function would induce leakage of urine from urethra, which means the reflux of urine to urethra is decreased. There were 45.9% of patients with UUTD showing normal urethra function, while none of the patients showing incompetent urethra function was with UUTD. The results indicated that urethra function during filling phase is a major predictive variable for risk of UUTD in NGB patients and the root node of the decision tree.

In our study, Pabd max and Pves max were valuable predictive variables for risk of UUTD in NGB patients. In normal conditions, under the effect of Pabd, a certain Pves in bladder is determined during the filling phase. When Pves in bladder is raised to a certain degree, micturition reflex is induced to maintain a lower pressure in bladder. As long as the intravesical pressure is in a safe range (less than 40 cm [H.sub.2]O) in the filling and voiding phases, urine does not reflux into ureter. If the intravesical pressure increases excessively, there is difficulty for urine in upper urinary tract to be delivered from ureter to bladder; then, the persistent high intravesical pressure would lead to UUTD (2,15-18). The present study showed that 57.4% of patients with Pabd max > 14 cm [H.sub.2]O were with UUTD, while 15% of patients with Pabd max [less than or equal to] 14 cm [H.sub.2]O were with UUTD, suggesting that high Pabd would increase the risk of developing neurogenic UUTD. In this study, there were 78.6% of patients with Pves max > 40 cm [H.sub.2]O . The results indicated that UUTD occurred in 79.2% of patients with Pves max [less than or equal to] 89 cm [H.sub.2]O, and among them there were many patients with Pves max > 40 cm [H.sub.2]O . Although some patients' Pves max were less than 89 cm [H.sub.2]O, the mean value of Pves max in each group was much higher than 40 cm [H.sub.2]O, which might indicate increased risk of neurogenic UUTD. The results also showed only 40% of patients with Pves max > 89 cm [H.sub.2]O developed UUTD. This indicates that the risk of UUTD gradually increases until the Pves reaches a certain degree and then the risk does not further increase when the Pves exceeds 89 cm [H.sub.2]O. The mean value of Pves max in each group was much higher than 40 cm [H.sub.2]O; however, because the sample size of our study was limited, it was quite difficult to find the exact cutoff point of Pves max to classify the risk of developing UUTD accurately. Further research and discussions with a larger sample size are needed.

The present study showed that gender is also a predictive variable for risk of UUTD in NGB patients. Because of different anatomical structures, women suffer more easily from urinary tract infection and high detrusor pressure, and the morbidity of UUTD in women may be higher than in men (3). The study revealed that gender might be a node in the decision tree, indicating that females had higher risks of neurogenic UUTD than males.

The decision tree model for risk of neurogenic UUTD established in this study summarized two classification rules. One is normal urethra function, Pabd max > 14 cm [H.sub.2]O, and Pves max [less than or equal to] 89 cm [H.sub.2]O, and the other is normal urethra function, Pabd max > 14 cm [H.sub.2]O, Pves max > 89 cm [H.sub.2]O and female gender. NGB patients in accordance with the classification rules may be at a higher risk of developing UUTD.

Our study developed a decision tree model to predict the risk of UUTD in patients with NGB. Further, it might provide a new and convenient way to screen UUTD in NGB patients in undeveloped and developing areas. 10.1590/1414-431X20176638


This work was supported by grants from the Natural Science Foundation of Guangdong Province China (No.2016A030313623), the Medical Research Project of Shenzhen China (No. 201601039), the Breeding Project of Clinical Research of Southern Medical University (No. LC2016PY060), and the Medical Research Foundation of Shenzhen Hospital, Southern Medical University. We thank all the participants in this study and all others who have provided us with support and help during the writing of this article. We thank Y. Chen from Foshan City First People's Hospital for assisting with data collection and J.J. Wyndaele from University of Antwerp and Antwerp University Hospital for assisting with manuscript preparation.


(1.) Sekerci CA, Isbilen B, Isman F, Akbal C, Simsek F, Tarcan T. Urinary NGF, TGF-beta1, TIMP-2 and bladderwall thickness predict neurourological findings in children with myelodysplasia. J Urol 2014; 191: 199-205, doi: 10.1016/j.juro.2013. 08.025.

(2.) Groen J, Pannek J, Castro Diaz D, Del Popolo G, Gross T, Hamid R, et al. Summary of European Association of Urology (EAU) Guidelines on Neuro-Urology. Eur Urol 2016; 69: 324-333, doi: 10.1016/j.eururo.2015.07.071.

(3.) Lee JS, Koo BI, Shin MJ, Chang JH, Kim SY, Ko HY. Differences in urodynamic variables for vesicoureteral reflux depending on the neurogenic bladder type. Ann Rehabil Med 2014; 38: 347-352, doi: 10.5535/arm.2014.38.3.347.

(4.) McGuire EJ, Woodside JR, Borden TA, Weiss RM. Prognostic value of urodynamic testing in myelodysplastic patients. J Urol 1981; 126: 205-209, doi: 10.1016/S0022-5347(17)54449-3.

(5.) Gerridzen RG, Thijssen AM, Dehoux E. Risk factors for upper tract deterioration in chronic spinal cord injury patients. J Urol 1992; 147: 416-418, doi: 10.1016/S0022-5347(17)37254-3.

(6.) Kurzrock EA, Polse S. Renal deterioration in myelodysplastic children: urodynamic evaluation and clinical correlates. J Urol 1998; 159: 1657-1661, doi: 10.1097/00005392-1998 05000-00084.

(7.) Weld KJ, Graney MJ, Dmochowski RR. Differences in bladder compliance with time and associations of bladder management with compliance in spinal cord injured patients. J Urol 2000;163: 1228-1233, doi: 10.1016/S0022-5347(05) 67730-0.

(8.) Habibi S, Ahmadi M, Alizadeh S. Type 2 diabetes mellitus screening and risk factors using decision tree: results of datamining. Glob J Health Sci 2015; 7: 304-310, doi: 10.5539/ gjhs.v7n5p304.

(9.) Brown E, Owen R, Harden F, Mengersen K, Oestreich K, Houghton W, et al. Predicting the need for adaptive radiotherapy in head and neck cancer. Radiother Oncol 2015, 116: 57-63, doi: 10.1016/j.radonc.2015.06.025.

(10.) Surucu M, Shah KK, Mescioglu I, Roeske JC, Small W Jr, Choi M, et al. Decision trees predicting tumor shrinkage for head and neck cancer: Implications for adaptive radiotherapy. Technol Cancer Res Treat 2016; 15: 139-145, doi: 10.1177/1533034615572638.

(11.) Omran DA, Awad AH, Mabrouk MA, Soliman AF, Aziz AO. Application of data mining techniques to explore predictors of HCC in Egyptian patients with HCV-related chronic liver disease. Asian Pac J Cancer Prev 2015; 16: 381-385, doi: 10.7314/APJCP.2015.16.1.381.

(12.) Lee BJ, Kim JY. Indicators of hypertriglyceridemia from anthropometric measures based on data mining. Comput Biol Med 2015; 57: 201-211, doi: 10.1016/j.compbiomed. 2014.12.005.

(13.) Brakeman P. Vesicoureteral reflux, reflux nephropathy, and end-stage renal disease. AdvUrol 2008; 508949, doi: 10.1155/2008/508949.

(14.) Radmayr C, Schwentner C, Lunacek A, Karatzas A, Oswald J. Embryology and anatomy of the vesicoureteric junction with special reference to the etiology of vesicoureteral reflux. TherAdv Urol 2009; 1: 243-250, doi: 10.1177/175628720 9348985.

(15.) Tekgul S, Riedmiller H, Hoebeke P, Kocvara R, Nijman RJ, Radmayr C, et al. European Association of Urology (EAU) guidelines on vesicoureteral reflux in children. Eur Urol 2012; 62: 534-542, doi: 10.1016/j.eururo.2012.05.059.

(16.) Gratzke C, Bachmann A, Descazeaud A, Drake MJ, Madersbacher S, Mamoulakis C, et al. EAU Guidelines on the assessment of non-neurogenic male lower urinary tract symptoms including benign prostatic obstruction. Eur Urol 2015; 67: 1099-1109, doi: 10.1016/j.eururo.2014.12.038.

(17.) Oelke M, Bachmann A, Descazeaud A, Emberton M, Gravas S, Michel MC, et al. EAU guidelines on the treatment and follow-up of non-neurogenic male lower urinary tract symptoms including benign prostatic obstruction. Eur Urol 2013; 64: 118-140, doi: 10.1016/j.eururo.2013. 03.004.

(18.) Stohrer M, Blok B, Castro-Diaz D, Chartier-Kastler E, Del Popolo G, Kramer G, et al. EAU guidelines on neurogenic lower urinary tract dysfunction. Eur Urol 2009; 56: 81-88, doi: 10.1016/j.eururo.2009.04.028.

H. Fang [1], B. Lu [1], X. Wang [2], L. Zheng [2], K. Sun [2] and W. Cai [1]

[1] Shenzhen Hospital, Southern Medical University, Shenzhen, China

[2] The Third Affiliated Hospital of Sun Yat-Sen University, Guangzhou, China

Correspondence: W. Cai: <>

Received May 3, 2017 | Accepted June 29, 2017

Caption: Figure 1. Decision tree model for risk of upper urinary tract damage (UUTD) in patients with neurogenic bladder. Pabd max: maximum abdominal pressure (> 14 cm [H.sub.2]O); Pves max (maximum intravesical pressure, [less than or equal to] 89 cm [H.sub.2]O),

Caption: Figure 2. Receiver operating characteristic analysis of the model, where the ROC curve is displayed in blue, and the reference is displayed in green.
Table 1. Baseline characteristics of neurogenic bladder patients
(NGB) with upper urinary tract damage (case group) or without it
(control group).

Characteristics                    Case group (n=34)

Age (years)                        49.9 [+ or -] 18.6
Gender (n, %)
  Male                                  17 (50%)
  Female                                17 (50%)
  NGB duration (days)                  7.5 (10.5)
Urinary tract infection (n, %)
  Yes                                  13 (38.2%)
  No                                   21 (61.8%)
Conservative treatment (n, %)
  No                                   21 (61.8%)
  Bladder function/pelvic              13 (38.2%)
    muscle exercise/
Bladder management (n, %)
  Urinate                               9 (26.4)
  Indwelling catheter/                 18 (52.9%)
  Clean intermittent                   7 (20.6%)
Detrusor activity (n, %)
  Normal                               21 (61.8%)
  Overactivity                         13 (38.2%)
Bladder sensation (n, %)
  Normal                               12 (35.3%)
  Increased                            10 (29.4%)
  Reduced/Absent                       12 (35.3%)
Bladder capacity (n, %)
  Normal                               12 (35.3%)
  High                                 10 (29.4%)
  Low                                  12 (35.3%)
Urethra function (n, %)
  Normal                               34 (100%)
  Incompetent                            0 (0%)
Detrusor contractility (n, %)
  Normal                               7 (20.6%)
  Underactive                          9 (26.5%)
  Acontractile                         18 (52.9%)
Bladder outlet obstruction
    (n, %)
  No                                   27 (79.4%)
  Yes                                  7 (20.6%)
Bladder compliance (n, %)
  [greater than or                     15 (44.1%)
    equal to] 20 cm
  <20 cm [H.sub.2]O                    19 (55.9%)
Maximal bladder
    capacity (n, %)
  [less than or equal                  10 (29.4%)
    to] 600 cm [H.sub.2]O
  < 600 cm [H.sub.2]O                  24 (70.6%)
Pdet max in filling
    phase (n, %)
  [less than or equal                  22 (64.7%)
    to] 40 cm [H.sub.2]O
  >40 cm [H.sub.2]O                    12 (35.3%)
Micturitional Pdet max (n, %)
  [less than or equal                  16 (47.1%)
    to] 40 cm [H.sub.2]O
  >40 cm [H.sub.2]O                    18 (52.9%)
Pves max in filling phase         74.14 [+ or -] 31.45
Pabd max in filling phase         26.08 [+ or -] 11.67

Characteristics                   Control group (n=78)     P

Age (years)                        45.2 [+ or -] 14.6     0.15
Gender (n, %)                                            0.205
  Male                                 49 (62.8%)
  Female                               29 (37.2%)
  NGB duration (days)                 15 (604.25)        0.023
Urinary tract infection (n, %)                           0.005
  Yes                                  52 (66.7%)
  No                                   26 (33.3%)
Conservative treatment (n, %)                            0.005
  No                                   26 (33.3%)
  Bladder function/pelvic              52 (66.7%)
    muscle exercise/
Bladder management (n, %)                                0.106
  Urinate                               18 (23)
  Indwelling catheter/                 54 (69.2%)
  Clean intermittent                    6 (7.7%)
Detrusor activity (n, %)                                 0.016
  Normal                               29 (37.2%)
  Overactivity                         49 (62.8%)
Bladder sensation (n, %)                                 0.004
  Normal                               11 (14.1%)
  Increased                            14 (17.9%)
  Reduced/Absent                       53 (67.9%)
Bladder capacity (n, %)                                  0.004
  Normal                               8 (10.3%)
  High                                 24 (30.8%)
  Low                                   46 (59%)
Urethra function (n, %)                                  <0.001
  Normal                               40 (51.3%)
  Incompetent                          38 (48.7%)
Detrusor contractility (n, %)                            0.014
  Normal                               10 (12.8%)
  Underactive                          44 (56.4%)
  Acontractile                         24 (30.8%)
Bladder outlet obstruction                               <0.001
    (n, %)
  No                                   34 (43.6%)
  Yes                                  44 (56.4%)
Bladder compliance (n, %)                                0.002
  [greater than or                     58 (74.4%)
    equal to] 20 cm
  <20 cm [H.sub.2]O                    20 (25.6%)
Maximal bladder                                          0.043
    capacity (n, %)
  [less than or equal                   39 (50%)
    to] 600 cm [H.sub.2]O
  < 600 cm [H.sub.2]O                   39 (50%)
Pdet max in filling                                      <0.001
    phase (n, %)
  [less than or equal                  15 (19.2%)
    to] 40 cm [H.sub.2]O
  >40 cm [H.sub.2]O                    63 (80.8%)
Micturitional Pdet max (n, %)                            <0.001
  [less than or equal                  11 (14.1%)
    to] 40 cm [H.sub.2]O
  >40 cm [H.sub.2]O                    67 (85.9%)
Pves max in filling phase         74.46 [+ or -] 35.29   0.964
Pabd max in filling phase         20.57 [+ or -] 13.13   0.037

Data are reported as means [+ or -] SD or number and percentages.
Pdet max: maximum detrusor pressure; Pabd max: maximum abdominal
pressure; Pves max: maximum intravesical pressure. Data were analyzed
with the t-test or chi-square test.

Table 2. Classification rules of the decision tree model for risk of
upper urinary tract damage (UUTD).

Rules                             Conditions

        Urethra     Pabd max             Pves max            Gender

1        Normal       >14       [less than or equal to] 89     --
2        Normal       >14                  >89               Female

Rules   Risk of UUTD   n

1          79.2%       19
2          58.8%       10

Pabd max: maximum abdominal pressure; Pves max: maximum intravesical

Table 3. Results of the decision tree model evaluation.

                                  Evaluation measures

          AUC     Sensitivity   Specificity   PV +    PV -    Accuracy

Results   0.901   85.3%         84.6%         70.7%   92.9%   84.8%

AUC: area under the curve; PV+: positive predictive value; PV-:
negative predictive value.

Table 4. Result of cross validation.

Method             Estimate   Standard error

Resubstitution      0.152         0.034
Cross-Validation    0.232         0.040
COPYRIGHT 2017 Associacao Brasileira de Divulgacao Cientifica (ABDC)
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2017 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Research Article
Author:Fang, H.; Lu, B.; Wang, X.; Zheng, L.; Sun, K.; Cai, W.
Publication:Brazilian Journal of Medical and Biological Research
Article Type:Report
Date:Oct 1, 2017
Previous Article:Antiviral activity of shikonin ester derivative PMM-034 against enterovirus 71 in vitro.
Next Article:Relationship between aerobic capacity and pelvic floor muscles function: a cross-sectional study.

Terms of use | Privacy policy | Copyright © 2022 Farlex, Inc. | Feedback | For webmasters |