Interpretation of pre- versus postimplant TRUS images.

In order to summarize the inter-observer variability of pre- and postimplant TRUS image interpretation. Ten patients treated with Pd-103 brachytherapy were studied. Preimplant prostatevolumes ranged from 21 to 51 cm3. The number of sources implanted ranged from 74 to 155, and the number of sources per cm3 prostate volume ranged from 3.0 to 4.3. A set of transverse images (6 MHz) were taken immediately prior to and following source placement. Original printer images were sent to four investigators and the prostate outlined independently on a cellophane overlay. The overlays were digitized into a Varian MMS 7.0 treatment planning system (Charlottesville, VA) for volume determinations. There was moderate interobserver variability in TRUS volume determination, accentuated for the postimplant images. The standard deviations varied from 2% to 13% of the mean (median: 7%) for preimplant volumes, versus 7% to 32% (median: 13%) for postimplant volumes. Interobserver prostatic edge (border) localization variability was greatest at the base and apex, with closer agreement along the posterior border. For preimplant images, the majority of edge points were within 1.0 mm of the mean. At each coordinate, with the exception of the anterior base, the majority of points were within 2.0 mm of the mean. In general, border identification variability was greater in the post implant images. While all prostate imaging modalities suffer from interobserver variability, preimplant and postimplant TRUS appears capable of consistently determining prostatic volume and borders. It appears that intraoperative TRUS-based dosimetry is a practical goal, provided that seed location coordinates can be added to the prostatic edge information derived from TRUS images.