Quantification of interobserver variability in image registration using cone beam CT for partial bladder...

The use of lipiodol or bladder wall surface (BWS) for image guidance has improved the treatment quality for partial bladder irradiation. Currently, this procedure is manually performed by different users. This study assessed the interobserver variability of using image guidance for partial bladder irradiation.

7 observers were prospectively recruited to manually register 5 cone beam CTs (CBCT) from each of 20 bladder cancer patients with lipiodol injected for tumor demarcation. Lipiodol and BWS were used to register the CBCT to pre-treatment reference images, and displacement values in three directions were collected. Mean difference among observers and the 95% limit of agreement were calculated to measure interobserver variability. Margin required and the resultant treatment volume were compared between the surrogates.

A total of 4200 displacement values were collected for analysis. Lipiodol was superior to BWS, with a mean difference among observers of <2 mm and a 95% limit of agreement of <5 mm in all directions. Of the three directions, greatest variability was observed in the superior-inferior direction for both surrogates, hence requiring a larger margin than the other two directions. After applying the corresponding margin, the mean volume of BWS-planning target volume was calculated to be significantly larger than lipiodol-planning target volume (166 cm³vs 134 cm³, p < 0.05).

The use of lipiodol achieved a higher interobserver agreement than BWS. A larger margin in the superior-inferior direction is recommended due to greater interobserver variability observed in this direction for both surrogates.

The uncertainty associated with the image registration by multiple observers for bladder image-guided radiotherapy is quantified for two surrogates.