Considerations for the clinical implementation of MRI-guided ART for H&N and lung cancers

Abstract

Purpose or Objective Clinical commissioning of complex treatment indications on the MR Linac (MRL) requires several considerations: selecting the optimal image sequences for planning and treatment delivery, ensuring users are competent in MR image registrations, adaptation strategy and identification of triggers for adaptation are identified. This work describes the steps taken to select MRI sequences and validate inter-observer registration prior to the clinical implementation of MR-guided adaptive radiotherapy for oropharyngeal (H&N) and lung cancers. Materials and Methods Prior to the clinical implementation on the MRL for H&N and lung cancer, patient volunteers were recruited to an institutionally approved imaging study (PRIMER). Image quality was assessed using visual guided assessments (VGAs) to determine the most suitable tissue weighting for daily image registration. Three observers (1 oncologist and 2 radiographers) independently scored visibility of the tumour and pre-determined organs at risk (OARs) on vendor provided T1 and T2 weighted images for 10 H&N and 6 lung patients. Nine radiographers completed offline MRI to CT and MRI to MRI rigid registrations in Monaco v5.11.02 (Elekta AB, Sweden) for 5 H&N and 4 lung patients, using bony and soft tissue matching strategies. The resulting translations, registration time and confidence scores were recorded using a 5 point Likert scale. Descriptive statistics were calculated in Microsoft Excel. Results The VGA scores demonstrated tumour visibility was unclear/not visible for all lung images, however, the lungs (44% vs 33%) bronchial tree (67% vs 44%) and trachea (89% vs 78%) were more visible on T2 than T1. In the R/L (-0.002 vs -0.15,) and A/P (0.05 vs 0.19) directions T2-CT mean inter-observer variation were lower than those for T1-CT for soft tissues matches. The mean time taken for offline T1-CT and T2-CT image registration was 4 minutes. These times are longer than expected for online registration which to date has averaged at 2:42 minutes. For H&N, T1 and T2 images have similar tumour visualisation (58% vs 53%). T1 sequences demonstrated superior nodal (61%vs 41%), optic nerve (100% vs 38%) and parotid (91% vs 44%) visualisation. Mean inter-observer variation and SD for T1-CT and T2-CT registration methods were greatest in the SI directions for both bony and soft tissue matches with the greatest variation present for T2-CT. Radiographers had a 75% agreement on whether to soft tissue or bony match on T1-CT image registration compared to 59% for T2-CT.The mean time was identical for the T1-CT and T2-CT at 5 minutes, however online this has been significantly reduced to an average of 2:16 minutes Conclusion Using systematic evaluations the interdisciplinary team was able to identify and agree on sequence selection for H&N (T1) and lung (T2). This process will be used moving forward to future MRL indications (e.g., pancreas). Work continues outside the vendor provided workflow to further optimise imaging.