Quantitative measurements of regional lung ventilation and perfusion are useful adjuncts to image interpretation.
This study investigated the accuracy and precision of the software used to carry out such measurements in the UK.
Ten examples of perfusion distribution, representing the range of patterns expected in practice, were simulated on computer using a segmental model of the lung and real three-dimensional lung shapes obtained from magnetic resonance images. Pairs of anterior and posterior perfusion images were simulated and distributed to UK hospitals wishing to take part in the audit. Each centre returned analysis results and technical details. Forty centres provided data on the relative right : left lung perfusion. Thirteen also submitted data with each lung divided into three zones and four with each lung divided into two zones. All measurements were expressed by the percentage of total perfusion in a particular region. Errors were assessed as the root-mean-square (rms) deviation from the true value.
Methods varied in the view used for analysis (80% geometric mean, 20% posterior) and the use of background subtraction (71% not used, 29% used). The rms error for percentage right assessment was 1.5 percentage points. This increased on two- and three-zone analysis to 3.8 and 4.3 percentage points, respectively. Differences in technique made little difference to whole-lung relative perfusion errors, but were important in zonal analysis.
Quantification of whole-lung relative function is accurate and reproducible. Zonal values are determined less accurately, but still provide a useful guide to the distribution of function.