Software output from semi-automated planimetry can underestimate intracerebral haemorrhage and peri-haematomal oedema volumes by up to 41.

ABSTRACT

INTRODUCTION: Haematoma and oedema size determines outcome after intracerebral haemorrhage (ICH), with each added 10 % volume increasing mortality by 5 %. We assessed the reliability of semi-automated computed tomography planimetry using Analyze and Osirix softwares. METHODS: We randomly selected 100 scans from 1329 ICH patients from two centres. We used Hounsfield Unit thresholds of 5-33 for oedema and 44-100 for ICH. Three raters segmented all scans using both softwares and 20 scans repeated for intra-rater reliability and segmentation timing. Volumes reported by Analyze and Osirix were compared to volume estimates calculated using the best practice method, taking effective individual slice thickness, i.e. voxel depth, into account. RESULTS: There was excellent overall inter-rater, intra-rater and inter-software reliability, all intraclass correlation coefficients >0.918. Analyze and Osirix produced similar haematoma (mean difference Analyze - Osirix = 1.5 ± 5.2 mL, 6 %, p ≤ 0.001) and oedema volumes (-0.6 ± 12.6 mL, -3 %, p = 0.377). Compared to a best practice approach to volume calculation, the automated haematoma volume output was 2.6 mL (-11 %) too small with Analyze and 4.0 mL (-18 %) too small with Osirix, whilst the oedema volumes were 2.5 mL (-12 %) and 5.5 mL (-25 %) too small, correspondingly. In scans with variable slice thickness, the volume underestimations were larger, -29%/-36 % for ICH and -29 %/-41 % for oedema. Mean segmentation times were 653 ± 402 min with Analyze and 906 ± 524 min with Osirix (p < 0.001). CONCLUSION: Our results demonstrate that the method used to determine voxel depth can influence the final volume output markedly. Results of clinical and collaborative studies need to be considered in the context of these methodological differences.