Computerized artifact detection and correction of uroflow curves: Towards a more consistent quantitative assessment of maximum flow

Objectives: To evaluate a computerized method of artifact detection and correction of uroflow and compare the quantitative assessment of maximum flow obtained by the computer with visual correction by experts. Methods: A total of 90 randomly chosen flows was scanned into the computer whereafter automated artifact detection and correction was performed according to pre-established rules implemented in the software. Three experts visually corrected the flows using the same artifact detection and correction specifications as the computer. Measuring agreement between different methods of assessment of maximum flow was evaluated by calculating the difference and the standard deviation (SD) of the differences. The repeatability of assessing the maximum flow value by the computer and by expert 1 was assessed by calculating the difference between 2 readings and the coefficient of repeatability. Results: The coefficient of repeatability of maximum flow after detection and correction of artifacts by the computer (0.38 ml/s) was slightly better when compared with the coefficient of repeatability between 2 observations by 1 expert (1.12 ml/s). The interobserver variation for the quantitative assessment of maximum flow appeared to be great. A total of 51% of the maximum flow values assessed by expert 2 was 1 ml/s or more greater than those assessed by expert 1. When comparing the results of the computer with those of the experts, the mean value of maximum flow from expert 1 was 0.71 ml/s smaller than the computer value (p < 0.01), the mean value from expert 2 was 0.53 ml/s greater (p < 0.01) and the mean value from expert 3 was not significantly different (0.25 ml/s greater). The SD of maximum flow after correction by the computer was 0.3 ml/s smaller than the SD of the raw data from the flowmeter and the corrected values by 2 experts. Conclusions: Computerized artifact detection and correction eliminates an important fraction of the variability of manually corrected maximum flow values. This may lead to smaller sample size requirements, especially in studies where the primary objective is to assess a small (+/- 1 ml/s) difference in mean maximum flow between groups.