Page 130 - 89Zr-Immuno-PET:Towards a Clinical Tool to Guide Antibody-based Therapy in Cancer

Chapter 6
When measurement variability for SUVmax was > 0, VOI were assessed for apparent insufficient tumor contrast for manual tumor delineation.
Based on this assessment VOI were deemed ineligible for quantification, according to the following criteria:
1. A different structure was delineated by at least 1 observer.
2. The voxel with maximum intensity was located at the border of
the VOI, of at least 1 observer.
Interobserver variability and reliability were analyzed for the entire group
of VOI, as well as for the subset of VOI eligible for quantification.
Interobserver reproducibility
Interobserver reproducibility for manual tumor delineation on immuno-PET was assessed by an agreement parameter (standard error of measurement (SEM)) as well as a reliability parameter (ICC; (11)). As we expected that the interobserver variability between lesions within a single patient was equal or higher than between patients, we performed a VOI-based analysis.
Interobserver variability. The agreement parameter reflects the measurement error due to interobserver variability (11). For every tumor lesion, three values (value1, value2 and value3) were obtained from observer 1, 2 and 3, respectively. Absolute interobserver variability was calculated as:
, [1]
where SD is the standard deviation.
SEM was calculated for each individual tumor lesion and has the same unit as the uptake measure (SUVmax, SUVpeak and SUVmean, dimensionless; volume in mL; TLU in %ID).
Relative interobserver variability was calculated as:
CoV = SEM / average (value 1, value 2, value 3) * 100, [2]
where CoV (%) is the coefficient of variation.
When all observers measure the exact same tumor uptake, SEM and CoV equal 0. Correlation of absolute and relative variability with tumor uptake was assessed.
128