Interobserver reliability for manual analysis of sidestream darkfield videomicroscopy clips after resizing in ImageJ

BackgroundDirect assessment of microcirculatory function remains a critical care research tool but approaches for analysis of microcirculatory videomicroscopy clips are shifting from manual to automated algorithms, with a view to clinical application in the intensive care unit. Automated analysis software associated with current sidestream darkfield videomicroscopy systems is demonstrably unreliable; therefore, semi-automated analysis of captured clips should be undertaken in older generations of software. We present a method for capture of microcirculatory clips using current version videomicroscope hardware and resizing of clips to allow compatibility with legacy analysis software. The interobserver reliability of this novel approach is examined, in addition to a comparison of this approach with the current generation of automated analysis software.ResultsResizing microcirculatory clips did not significantly change image quality. Assessment of bias between observers for manual analysis of resized clips; and between manually analysed clips and automated software analysis was undertaken by Bland-Altman analysis. Bias was demonstrated for all parameters for manual analysis of resized clips (total vessel density = 6.8, perfused vessel density = 6.3, proportion of perfused vessels = - 8.79, microvascular flow index = - 0.08). Marked bias between manual analysis and automated analysis was also evident (total vessel density = 16.6, perfused vessel density = 16.0, proportion of perfused vessels = 1.8). The difference between manual and automated analysis was linearly related to the magnitude of the measured parameter.ConclusionsPoor reliability of automated analysis is a significant hurdle for clinical translation of microcirculatory monitoring. The method presented here allows capture of microcirculatory clips using current hardware that are backwards compatible with older versions of manual analysis software. We conclude that this approach is appropriate for research applications in the intensive care unit, however the time delay to results limits utility for clinical translation.

Automated summary

The study introduces a method for capturing microcirculatory clips using current hardware that are compatible with older versions of manual analysis software. Results showed that resizing clips did not significantly affect image quality. There was marked bias between manual and automated analysis, with a linear relationship between the differences in the analyses and the magnitude of the measured parameter.