ESTRO 2021 Abstract Book

S145

ESTRO 2021

average motion in the test-retest had a vector length of 1.2cm (0.4-3.3cm). A moderate correlation between motion amplitude and ADC difference was observed for both observers (R = 0.43 and 0.45)(Fig.2).

Conclusion The ADC values for liver metastases measured on a MR-linac were in the same range as ADC values reported on diagnostic systems 3,4,5 . The %RC values for this dataset are larger than those reported for free-breathing DWI on diagnostic systems (26% 2 and 30% 3 ). The inter-observer variation was smaller than the variation between the test and retest measurements, suggesting that the test-retest variation is not dominated by delineation variation. We also did not find a strong relation with the amount of tumor motion. However, the ADC values might be affected by misalignment between individual b-value images. The repeatability could be improved by correction for motion by triggered acquisitions or retrospective registration 3 . References

1. Padgett et al. Phys. Med. 2020 2. Shukla-dave et al. JMRI 2019 3. Pathak et al. Sci. Rep. 2019 4. Deckers et al. JMRI 2014 5. Lahrsow et al. CIRSE 2017

PH-0214 Does longitudinal Diffusion-Weighted MRI have the potential to carry biological information? A. Bisgaard 1 , C. Brink 1,2 , M. Lynge Fransen 3 , T. Schytte 1,4 , C. Behrens 5 , H. Nissen 6 , F. Mahmood 1,2 1 Odense University Hospital, Department of Oncology, Odense, Denmark; 2 University of Southern Denmark, Department of Clinical Research, Odense, Denmark; 3 Odense University Hospital, Department of Radiology, Odense, Denmark; 4 University of Southern Denmark, Department of Clinical Research, Odense , Denmark; 5 Herlev Hospital, Department of Oncology, Herlev, Denmark; 6 Vejle Hospital, Department of Oncology, Vejle, Denmark Purpose or Objective Introduction of the hybrid MRI linear accelerator (MR-linac) has made longitudinal Diffusion-Weighted MR imaging (DWI) more accessible. This allows studying the temporal changes of quantitative DWI metrics such as Apparent Diffusion Coefficient (ADC), a promising biomarker for response prediction. ADC measurement requires delineation of ROIs, which is time-consuming and can be error-prone. Here, a, semi-automatic computer-based tool for segmentation of dedicated ROIs ( viable tumor volumes , VTV) for ADC measurement is tested for its capacity to detect potential biological changes. Materials and Methods A semi-automatic segmentation tool was implemented using in-house developed software (MatlabR2019a, Mathworks ab, Sweden), as a 3-step process (Figure 1): 1) Manual input for identifying roughly the relevant region. 2) Two masks are automatically created with high DWI intensity and low ADC values, respectively, based on Otsu’s method to identify discrimination thresholds 1 . 3) The overlap between the two masks form the resulting VTV. The tool was tested in 30 patients with rectal cancer referred to RT and MRI scanned prospectively before treatment (baseline) and two weeks into RT (week 2). MRI protocol consisted of T2W imaging and repeated DWI (test-retest). Rigid registration between T2W and DWI was performed in MIM (MIM Software Inc.). A