INTEROBSERVER VARIABILITY IN ROI DRAWING IN COMPUTERIZED SCAN PROCESSING

Backgroud. Nuclear medicine uses computerized scintillation cameras fo r scan processing. For the radiorenogram, the main objective is to obs erve the passage of radionuclides through the kidneys during a given t ime. The new software automatically provides data of the dynamic studi es, but there is one step (the drawing of regions of interest, ROI) ha ndled by an operator. With the ROI drawings the computer integrates th eir radioactivity and displays it as time/activity curves, and calcula tes the T-max (time to achieve maximal activity) and the T 1/2 (time t o eliminate half of the T-max) of each kidney. Objective. To evaluate observer intervariability in drawing the ROI on renal scans. Material and methods. Four observers with at least seven years of experience in the prodedure drew independently the ROI of 38 renograms of 20 patien ts (two transplants) to obtain the T-max and T 1/2. The interobserver CV was calculated for the T-max and the T 1/2 of the 38 scans. Results . Globally the interobserver variability was larger for T 1/2 than for T-max. There were four scans with small differences in Tmax and/or T 1/2 but which lead to interobservers discrepancies in the classificati on (normal/abnormal). The partition of the scans in three groups (1 - T-max and T 1/2 normal; 2 = only T-max normal; 3 = both abnormal) show ed significant intergroup differences in the interobserver variability (Kruskal-Wallis p = 0.032) which were caused by the larger variabilit y in group 2 (6 of 11 scans with CV > 4%) than in the other groups (no ne with CV > 4%). Conclusions. 1. The interobserver discrepancies in c lassification were observed only in cases with parameters slightly abn ormal (8-9 min in T-max 15-20 min in T 1/2). 2. We have no explanation for the larger interobserver variability of the T-max in group 2 (nor mal T-max with abnormal T 1/2). 3. We believe a study of intraobserver variability in these same four observers may help in gaining insight to some of the observations in this study.