ESTRO 2021 Abstract Book

S147

ESTRO 2021

1. Otsu N. Threshold Selection Method From Gray-Level Histograms. IEEE Trans Syst Man Cybern . 1979;SMC-9(1):62-66. doi:10.1109/tsmc.1979.4310076

PH-0215 Repeatability of diffusion-weighted MRI in head and neck cancer at a 1.5 T MR-Linac J. Habrich 1 , S. Boeke 2 , M. Nachbar 1 , K. Nikolaou 3 , F. Schick 4 , C. Gani 2 , D. Zips 2 , D. Thorwarth 1 1 University of Tübingen, Section for Biomedical Physics, Department of Radiation Oncology, Tübingen, Germany; 2 University Hospital and Medical Faculty, University of Tübingen, Department of Radiation Oncology, Tübingen, Germany; 3 University Hospital Tübingen, Department of Radiology, Tübingen, Germany; 4 University of Tübingen, Section for Experimental Radiology, Department of Diagnostic Radiology, Tübingen, Germany Purpose or Objective With the clinical introduction of hybrid MR-Linacs (MRL), sequential assessment of functional tumor characteristics during radiotherapy (RT) seems highly promising with respect to biologically adapted RT approaches. Studies using offline MRI suggest a prognostic value of apparent diffusion coefficients (ADC) assessed via diffusion weighted (DW) MRI in head and neck cancer (HNC) RT. Before clinical validation of ADC as imaging biomarker in online MR-adaptive RT, a dedicated technical validation is required. The aim of this study was to determine the repeatability of ADC values derived from DW-MRI in patients with HNC acquired at a hybrid 1.5 T MRL. Materials and Methods Six patients with locally advanced HNC were treated at the 1.5 T MRL and investigated with an extended imaging protocol (ethics approval No. 659/2017BO1 ) . Sequential test/re-test measurements using an echo- planar-imaging (EPI) DW-MRI technique with b-values of 0, 150, 500 s/mm 2 were performed once a week before (test) and after (re-test) the treatment session. ADC maps were calculated using a pixel-wise mono- exponential fit to signal amplitudes of the b150- and b500-images. Volumes of interest (VOI) including the gross tumor volume (GTV), all suspicious lymph nodes (LN), both submandibular and parotid glands were delineated manually on the b500-image. Voxel-based analysis was carried out for a voxelsize of 4.2x4.2x4 mm 3 . Repeatability coefficient (RC) as well as within-subject coefficient of variation (wCV) with respect to ADC were calculated on VOI- and voxel-level. Bland-Altman analysis was used for VOI-based results. Results A total of 40 test/retest measurements were available for analysis. Mean ADC values for GTV, LN, left/right submandibular and parotid gland were 1370, 1152, 1270/1228 and 792/837 ∙10 -6 mm 2 /s, respectively. On the VOI-level wCV ranged between 3.3% in the parotid glands and 8.3% in the GTV. Accordingly, RCs in the range 81-335∙10 -6 mm 2 /s were observed. On a voxel-level, a mean wCV of 28.4% over all VOIs was determined. RC and wCV for VOI- and voxel-based evaluation are presented in Table 1. Figure 1 shows Bland-Altman plots for all six VOIs. Mean ADC differences for GTV, LN, left/right submandibular and parotid glands were -92.2, -35.5, -73.5/-42.1, -32.0/-36.8 ∙10 -6 mm 2 /s, respectively.

Submand. left Submand. right Parotid left Parotid right GTV LN

RC [∙10 -6 mm 2 /s] 217.5

VOI

221.2

81.0

95.1

334.5 177.8

wCV [%]

6.2

7.1

3.3

3.8

8.3 5.5

Voxel-level RC [∙10 -6 mm 2 /s] 695.7

734.7

541.1

548.3

939.1 797.3

wCV [%]

23.6

26.4

32.4

29.7

29.1 29.4