Role of artificial neural network and logistic regression model in predicting effect of extracorporeal shock wave for upper urinary tract calculi

Abstract

Objective To explore the role of artificial neural network and logistic regression model in predicting the effect of extracorporeal shock wave for upper urinary tract calculi. Methods From January, 2010 to January, 2015, d 340 patients with renal calculus were treated by ESWL at our hospital. The predictive parameters were sex, symptoms induced by urethral irritation, blood urine, renal colic, stone position, stone of one side, age, BMI, disease course, and stone size. Artificial neural network and logistic regression model were built basing on these parameters to predict the clinical effect of ESWL for calculus of upper urinary tract. Results The most important five parameters in artificial neural network were stone size, disease course, blood urine, stone position, and BMI, with statistical differences (P<0.05). The most important parameters in logistic regression model were disease course, blood urine, and stone position, with statistical differences (P<0.05). Conclusions Artificial neural network and logistic regression model in predicting the effect of extracorporeal shock wave for upper urinary tract calculi are both highly accurate, so both are worth being clinically generalized. Key words: Artificial neural network; Logistic regression model; Upper urinary tract calculi; Extracorporeal shock wave