In this study, a prototype artificial neural network model (ANN) was used to estimate the stone passage rate and to determine the effectivity of predictive factors on this rate in… Click to show full abstract
In this study, a prototype artificial neural network model (ANN) was used to estimate the stone passage rate and to determine the effectivity of predictive factors on this rate in patients with ureteral stones. The retrospective study included a total of 192 patients with ureteral stones, comprising 128 (66.7%) men and 64 (33.3%) women. Patients were divided into two groups. Group 1 (n: 125) consisted of people who spontaneously passed their stones, Group 2 (n: 67) consisted of people who could not pass stones spontaneously. The groups were compared with regard to the relationship between input data and stone passage rate by using both ANN and standard statistical tests. To implement the ANN, the patients were randomly divided into three groups: (a) training group (n = 132), (b) validation group (n = 30), and (c) test group (n = 30). The accuracy rate of ANN in the estimation of the stone passage ratio was 99.1% in the group a, 89.9% in the group b, and 87.3% in the group c. It was revealed that certain criteria (stone size, body weight, pain score, ESR, and CRP) were relatively more significant for saving treatment cost and time and for avoiding unnecessary treatment. ANN can be highly useful for the avoidance of unnecessary interventions in patients with ureteral stones as it showed remarkably high performance in the estimation of stone passage rate (99.16%).
               
Click one of the above tabs to view related content.