Back to 2014 Fall Congress Meeting Abstracts

BACKGROUND
To evaluate the accuracy of a flowmeter coupled to a neuronal network that has learned to analyze uroflow signals giving spoken biofeedback to children with dysfunctional voiding
METHODS
A computerized neural network has learned to analyze a flow signal into 4 categories on a 5 point scale. Categories are interrupted flow, staccato flow, obstruction and flow time. Coupled to interrupted and staccato flow the system has been taught to give spoken biofeedback to the patient, such as: please do not push, please relax better or well done. To teach the system 400 uroflow curves have been judged by a team of experts and fed into the system. This study was done to test the accuracy of the system. 100 new uroflows have been judged by the system and compared to the judgment of the expert team
RESULTS
Agreement between the system and the team, maximal 2 points of difference on the 5 point scale, was 96%. Full agreement of the team with the spoken text was 68%. In 5% of flows, the system overestimated the quality with false-positive comment. In 27% the system overrated the amount of staccato due to irregular vibrations in the curve with negative feedback to the patient where well done would have been appropriate
CONCLUSIONS
We have constructed a fully automated flowmeter for biofeedback training. Some fine tuning in the aspect of staccato has been done and the system is now incorporated in a clinical study that will be presented in the next years