Non-invasive Parameters Prediction Of Bladder Outlet Obstruction Of Benign Prostatic Hyperplasia Patients

Objective:Benign Prostatic Hyperplasia(BPH) is a very common disease in elderly men with a highly prevalence, when bladder outlet occur in obstruction, the patient will show some lower urinary tract symptoms, affecting the quality of life of patients seriously. There is no doubt that for patients with bladder outlet obstruction, surgery to remove the obstruction is the most ideal method of treatment, but urination disorders in some patients with BPH is not entirely caused by bladder outlet obstruction. Therefore, it is very important for BPH patients to identify bladder outlet obstruction. This study adopts a series of non-invasive parameters and a new prediction model was established by using the artificial neural network model, in order to predict bladder outlet obstruction better.Materials and Methods:We evaluated 200 cases of patients with BPH of Shanxi Dayi Hospital from February2013 to March 2015, include 160 cases as training group and 40 cases as validation group.A three layer BP neural network is constructed by seven non-invasive parameters, we use160 cases of patients to establish a network model, and through the validation group of 40 patients to test the performance of the model.Results:Based on age, IPSS, volume of prostate, maximum urinary flow rate(Qmax), post voiding residual, IPP and whether urinary retention, we set up the neural network model,the model have a nice predictive value, with an average error of 0.1966, and predicted performance of 80.34%. For the validation set, when A-G cut off value is 40 cm H2 O,the sensitivity and specificity of neural network prediction model is 75% and 58%,respectively.Conclusion:This study suggests that the artificial neural network can better explain the interaction of the nonlinear relationship between variables compared with the traditional regression model, so it can improve the prediction of BOO, but still not enough to completely replace the urodynamic results. Therefore, patient with LUTS wants to get definite diagnosis, he need to do urodynamic exam, at the same time, BP neural network can identify which patients must accept PFS, and which part of the patient can pass PFS, so as to reduce the burden of patients.