Comparison of prostate volume, shape, and contouring variability determined from preimplant magnetic resonance and transrectal ultrasound images

Abstract: Purpose: To compare preimplant prostate contours and contouring variability between magnetic resonance (MR) and transrectal ultrasound images. Methods and Materials: Twenty-three patients were imaged using ultrasound (US) and MR before permanent brachytherapy treatment. Images were anonymized, randomized, and duplicated, and the prostate was independently delineated by five radiation oncologists. Contours were compared in terms of volume, dimensions, posterior rectal indentation, and observer variability. The Jaccard index quantified spatial overlap between contours from duplicated images. Results: The mean US/MR volume ratio was 0.99±0.08 (p =0.5). The width, height, and length ratios for the prostate were 0.98±0.06 (p =0.09), 0.99±0.08 (p =0.4), and 1.05±0.14 (p =0.1). Rectal indentation was larger on US by 0.18mL (p =0.01) and correlated with prostate volume (p <0.01). MR and US interobserver variability in volume were similar at 3.5±1.7 and 3.3±1.9mL (p =0.6). Intraobserver variability was smaller on US at 1.4±1.1mL compared with MR at 2.4±2.2mL (p =0.01). Local intraobserver variability was lower on US at the midgland slice (p <0.01) but lower on MR at the base (p <0.01) and apex (p <0.01) slices. Conclusions: US is comparable to MR for preimplant prostate delineation, with no significant difference in volume and dimensions. Rectal indentation because of the transrectal ultrasound probe was measurable, although the effects were small. Intraobserver variability was lower on US for the prostate volume but was lower on MR locally at the base and apex. However, the difference was not observed for the interobserver variability, which was similar between MR and US. [Copyright &y& Elsevier]