Comparison of prostate delineation on multimodality imaging for MR-guided radiotherapy

OBJECTIVE: With increasing incorporation of MRI in radiotherapy, we investigate two MRI sequences for prostate delineation in radiographer-led image guidance. METHODS: Five therapeutic radiographers contoured the prostate individually on CT, T(2) weighted (T(2)W) and T(2)* weighted (T(2)*W) imaging for 10 patients. Contours were analysed with Monaco ADMIRE (research v. 2.0) to assess interobserver variability and accuracy by comparison with a gold standard clinician contour. Observers recorded time taken for contouring and scored image quality and confidence in contouring. RESULTS: There is good agreement when comparing radiographer contours to the gold-standard for all three imaging types with Dice similarity co-efficient 0.91–0.94, Cohen’s κ 0.85–0.91, Hausdorff distance 4.6–7.6 mm and mean distance between contours 0.9–1.2 mm. In addition, there is good concordance between radiographers across all imaging modalities. Both T(2)W and T(2)*W MRI show reduced interobserver variability and improved accuracy compared to CT, this was statistically significant for T(2)*W imaging compared to CT across all four comparison metrics. Comparing MRI sequences reveals significantly reduced interobserver variability and significantly improved accuracy on T(2)*W compared to T(2)W MRI for DSC and Cohen’s κ. Both MRI sequences scored significantly higher compared to CT for image quality and confidence in contouring, particularly T(2)*W. This was also reflected in the shorter time for contouring, measuring 15.4, 9.6 and 9.8 min for CT, T(2)W and T(2)*W MRI respectively. Conclusion: Therapeutic radiographer prostate contours are more accurate, show less interobserver variability and are more confidently and quickly outlined on MRI compared to CT, particularly using T(2)*W MRI. Advances in knowledge: Our work is relevant for MRI sequence choice and development of the roles of the interprofessional team in the advancement of MRI-guided radiotherapy.