Image registration guidelines impact on radiographer inter-observer variability for cervical cancer

Abstract

Purpose or Objective This group has previously reported the impact of an educational tool on MR-CT inter-observer image registration variation using a cohort of 4 radiographers (RTI). The introduction of an MR-MR workflow required an extension of this work to evaluate the impact of experience with MR guided Radiotherapy (MRgRT) compared with the use of an image registration guide across three imaging modalities, CBCT-CT, MR-CT and MR-MR. Materials and Methods Eight radiographers participated and were divided into two cohorts, 4 with experience in MRgRT and Monaco planning system (RTMR), and 4 with limited MRgRT and Monaco experience (RTCT). Observers completed offline retrospective image registration in Monaco (v5.11.02) for 10 patients with cervical cancer including 3 CBCT-CT, 3 MR-CT and 2 MR-MR registrations per patient. All observers received an initial overview of Monaco functionality, and an introduction to MRI-based gynaecological anatomy. Rigid translational registrations were undertaken independently using soft tissue alignment of the cervix in all three planes. Resulting translations, time taken for registration and confidence scores using Likert scale of 1(not confident) to 5 (extremely confident) were recorded and compared between cohorts (RTMR, RTCT, RTI) using descriptive statistics. Following a 14 day break, the process was repeated with an education guide (developed by a multi-disciplinary team including therapeutic/diagnostic radiographers and clinicians) being introduced aiming to standardize image registration by highlighting the most relevant anatomic regions. Results Collation of the post-intervention registrations for RTMR and RTCT cohorts is on on-going. However, the initial mean registration confidence was higher in the RTMR group for MR-CT and MR-MR registrations, 2.8 and 3.2 respectively compared to the RTCT group 2.6 and 2.8. The RTCT group were more confident with CBCT-CT registrations (mean 2.1(RTCT) vs 1.7 (RTMR )). The mean pre-intervention MR-CT registration confidence levels for RTI are similar to RTMR (3.2 vs 2.8) with the RTCT cohort slightly lower (2.6). There was no statistically significant difference in registration time between all cohorts and all registration methods, with combined group mean registration times of 2.3 (1.0, 2-5), 2.4 (1.9, 1-10), 2.1(0.6, 1-4) minutes for CBCT-CT, MR-CT and MR-MR respectively. Variation in inter-observer registration was smallest in like-for-like image registration methodologies (CBCT-CT and MR-MR) for the RTMR cohort and in the CT guided methodologies in the RTCT cohort. Table 1 summarizes pre intervention inter-observer registration variations for the RTCT and RTMR cohorts. Conclusion Based on preliminary data it is anticipated that the combination of both a dedicated image registration guide and familiarity with MRI and the MRL specific software will reduce inter-observer variation and time while increasing confidence in undertaking at-treatment MR-CT and MR-MR registrations.