It has been reported that urinary leakage was not correlate to urethral pressure profile (UPP) including maximum urethral closure pressure (MUCP).　This study aimed to investigate the factors related Valsalva leakage (VL) and MUCP using decision tree through machine learning.

24 patients who underwent urodynamics including Valsalva leakage point pressure (VLPP) and UPP measurement at our hospital from 2017 to 2020 were recruited. We described VL (obtained from VLPP measurement) and MUCP (obtained from UPP measurement) as objective variables using decision tree ( non- parametric supervised learning algorithm) through machine learning. As explanatory variable, particularly quantitative variables, age, MUCP, functional urethral length (FUL), vesical pressure at Valsalva maneuver voluntarily and using syringe, delivery times, body mass index (BMI), weight, height were defined. We set one and five as a minimum leaf size according to optimizing leaf size analysis and 10 as cross validation test times. Calculation in decision tree were performed using MATLAB™（Math Works, Natick, US）; mean squared error (the average squared error between actual and predicted values: MSE) was measured the quality for the estimator of the decision tree.

Four patients (20.8%) were detected VL; MUCP was 41.5 (31.75 – 60.75). FUL (28.95mm) and weight were plotted at decision tree for presence of VL as objective variable, importance was 0.0228 and 0.0198 respectively. After analysis, MUCP at a leaf detected VL was 36 (31 - 67) cmH2O and at the other leaves not detected VL were 44 (31.5 – 60.75: FUL≧28.95) cmH2O, 45 (34.5 – 59.5: Weight = 60) cmH2O, and MSE was 0.0417. Age (75.5 years old) and FUL (29.25mm) were described on decision tree for MUCP as objective variable, importance was 0.0349,0.0273 respectively. After analysis, the maximum detection rate was 80% (4 / 5 patients) at a leaf classified MUCP and 20% (1 / 5 patient) at the other leaf, and MSE was 575.29.

In this study, we investigate the factors related Valsalva leakage (VL) and maximum closure urethral pressure (MUCP) using decision tree through machine learning. It has known that urinary leakage was not correlate to UPP including MUCP in clinical practice. Although, it might be beneficial to understand the relationship urinary leakage and UPP including MUCP, obtained objective and explanatory variables from the analysis of VL and MUCP related factors using decision tree through machine learning.

These results suggested that the analysis of factors related VL and MUCP using decision tree through machine learning is beneficial to understanding the relationship urinary leakage (VL) and UPP including MUCP