Purpose or Case Report: To determine the inter-observer agreement in measurement of relative cerebral blood volume (rCBV) in pediatric brain tumors, using normal grey matter as a control.
Methods & Materials: Thirty one pediatric patients with malignant brain tumours (WHO grade 2-4) who had T2* perfusion weighted imaging performed at initial presentation were included. All examinations were performed on a 1.5 T GE Cigna magnet. A pediatric radiologist and fellow independently assessed the highest rCBV in the tumor as well as the normal caudate head and thalamus using the GE AW4 workstation and Functool software. Inter-observer agreement was assessed using Bland Altman plots.

Results: There was relatively good agreement in assessment of rCBV within the tumors with a mean bias of only 3.9% (95% CI -3.2% to 11.0%) and a standard deviation ( SD) of 19.2%. The upper 95% level of agreement (LOA) was 34.2% (95% CI 22.0 to 46.4%) and the lower 95% LOA was -42.0% (95% CI -54.2 to -29.8 %). Similar results were obtained for the caudate head rCBV, mean bias was 4.0% (95% CI -1.8% to 10.0%), SD 16.0% with upper LOA of 35.6% (95% CI 25.5 to 45.7%) and lower LOA of -27.4% (95% CI -37.4 to -17.3%). For the thalamus the mean bias was 7.0%, SD 19.8%, with upper LOA of 45.9% ( 95% CI 33.5 to 58.3%) and lower LOA of -31.5% ( 95% CI -43.9 to -19.1%).

Conclusions: While mean bias between observers was relatively low for both tumor and grey matter rCBV the 95% LOAs ranged between -27.4% and 45.9%, implying at best moderate agreement which may limit the clinical utility of T2* perfusion. Further work is required to establish whether this level of agreement is acceptable for clinical use and how it could be improved.