Purpose: Pre- and post-treatment MRI based Diffusion Weighted Imaging (DWI) has shown to be an effective predictor and indicator of treatment response in cancer. However, clinical applications of the Apparent Diffusion Coefficient (ADC) requires sufficient precision; this can be assessed with repeatability studies. Therefore, the purpose of this study was to assess the repeatability of ADC measurements on an MR-Linac for patients with locally advanced rectal cancer (LARC). Additionally, the relative change in ADC during treatment was compared against the repeatability.
Methods: For 17 patients, DWI was performed once on 3T MRI during pre-treatment, and twice on an 1.5T MR-Linac during each treatment fraction. Manual delineations of the Gross Tumour Volume (GTV) were created by the author. In addition two semi-automatic delineation methods were implemented, which used; a registration pipeline to propagate T2 delineations; a geometric distortion correction algorithm to correct for DWI susceptibility artefacts. Based on visual inspection the most accurate and consistent delineations were used to calculate the ADC; Bland-Altman repeatability coefficient (RC), within subject coefficient of variation (wCV), and the intraclass correlation coefficient (ICC). Lastly, the relative change in mean and median tumour ADC was compared against the RC.
Results: Manual delineations were determined to have the highest agreement with ADC tumour location and were used in subsequent calculations. The mean and median tumour ADC wCV was 7.6% (CI95:5.5-9.2%) and 8.2% (CI95:5.9-9.9%) respectively, which corresponds with a RC of 21.0% (CI95:15.1-25.6%) and 22.7% (CI95:16.3-27.4%). The reliability of the measured data was good, with an ICC of 0.81 (CI95:0.72-0.87). In 6 out of 17 patients the relative change in ADC exceeded the RC at some point during treatment, which suggest a potential for future clinical applications.
Conclusion: Despite the low precision of rectal cancer ADC measurements, daily DWI imaging on MR-Linac has been shown to be viable for clinical applications in a subset of patients whom show significant ADC changes during treatment. Further studies are required to determine whether the measured ADC repeatability allows for clinically relevant observations.

Purpose:The MR-linac is a novel hybrid system that is used in radiotherapy (RT) and combines irradiation with MR imaging. Novel adaptive image-guided radiotherapy techniques are developed for this machine. However, new qualityassurance (QA) techniques are required that check if the (adapted) planned radiation dose is the same as the dose given to the patient. Dose reconstruction, based on the MR-LINACs logfile, can be a new powerful QA tool. Here, we verify logfile based dose reconstruction with the treatment planning system (TPS) dose and Delta4 phantom MR+ on various timescale and discuss its potential as a new QA tool for the MR-LINAC.Method:Software was developed and validated for comparison of the different dose distributions. A patient RT-plan was selected with the intensity modulated RT step-and-shoot technique. The first experiment TPS dose – Logfile was performed at timescales: 1. Total RT-plan 2. Per beam. The second experiment Delta4 – Logfile was performed at timescales: 1. Total RT-plan 2. Per beam 3. Per segment 4. Per 200ms.For analysis, visual comparison, dose profiles, dose difference (DD) and gamma-index (90% of the data) were used.Results:The patient RT-plan had 7 fields and 57 segments. Outcomes of the first experiment were: 1. Total RT-plan: DD = -1.5 to 1.5% and gamma-index (DD=1% & DTA=1mm) of 0.89. 2. Per beam: DD = -3.3 to 2.4% and gamma-index (DD=2% & DTA=2mm) of 0.87.Outcomes of the second experiment were: 1. Total RT-plan: DD of -2.7 to 2.4% and a gamma-index (DD=2% and DTA=2mm) of 0.69. 2. Per beam: maximum DD of -4.7 to 2.5% and gamma-index (DD=2% and DTA=2mm) of 0.99considering all beams. 3. Per segment: maximum DD of -5.9 to 6.0% and gamma-index (DD=3% and DTA=3mm) of 1.0 considering all segments. 4. Per 200ms: the cumulative DD in a high irradiated area was -4.0 to 7.1%, after the2nd segment and -5.2 to 5.3%,. after the 7th segment.Conclusion:Logfile based dose reconstruction might be newest QA tool for the MR-LINAC. Accuracies meet current accepted criteria for IMRT treatment plans considering the total RT-plan and per beam timescale. Smaller timescales have a slightly higher error and require further optimization.