ESTRO 2021 Abstract Book

S70

ESTRO 2021

0.18). A statistically significant difference was seen between ADC min

for malignant (0.42 ± 0.30 x 10 ¯³ mm ² /s)

vs. benign (0.72 ± 0.37 x 10 ¯³ mm ² /s) nodes (p = 0.02). The ROC analysis indicated that an ADC mean threshold < 1.49 x 10 ¯³ mm ² /s detected disease with 86% sensitivity, 33% specificity, and 59% accuracy. Whilst an ADC min threshold < 0.69 x 10 ¯³ mm ² /s had 86% sensitivity, 60% specificity, and 72% accuracy (Fig.1).

Conclusion ADC min shows value in predicting persistent lymph node malignancy post-(C)RT in primary head and neck cancer and appears more discriminatory than ADC mean . Further prospective studies in a larger sample size are warranted. OC-0096 Does inter-observer variation affect functional imaging quantification in DW-MRI for HNSCC? L. Hay 1 , I. McCrea 2 , A. Duffton 3 , P. McLoone 4 , C. Paterson 5 1 The Beatson West of Scotland Cancer Centre, Radiotherapy, Glasgow, United Kingdom; 2 Queen Elizabeth University Hospital , Department of Radiology, Glasgow, United Kingdom; 3 The Beatson West of Scotland Cancer Centre , Radiotherapy, Glasgow, United Kingdom; 4 Glasgow University , Health and Wellbeing, Glasgow, United Kingdom; 5 The Beatson West of Scotland Cancer Centre, Oncology , Glasgow, United Kingdom Purpose or Objective Apparent diffusion coefficient (ADC) measured using diffusion-weighted (DW)-MRI before and during radiotherapy (RT) is under investigation as an imaging biomarker. This may allow delivery of personalised, biologically adaptive radiotherapy for head and neck cancer (HNC) 1 . Although inter-observer variation (IOV) is well documented in RT target delineation, the impact of IOV when measuring ADC to quantify treatment response is not previously described. The aim of this work was to evaluate IOV in delineated target volumes between a RTT and the gold standard to determine if IOV affects measured ADC on the ADC map generated from DW-MRI. Materials and Methods MRI scans were acquired at baseline (MRI_1) and #10 (MRI_2) of radical RT as part of the MeRInO study 1 . Gross target volume (GTV) delineation and ADC measurement of primary (GTVp) and lymph node (GTVn) disease was undertaken on both scans within the contouring workspace of the Eclipse TM treatment planning system, as previously described 2 . The structures were manually delineated independently by the RTT and clinical oncologist (CO) on the T1 post contrast fat saturated (T1PCFS) sequences. The gold standard was a delineation of each target volume by the CO amended or verified by the radiologist (RO). Mean differences and standard deviations of all the GTVs contoured were measured between observers. Inter- observer variation on the T1PCFS sequence was evaluated using paired t-tests and repeated measures ANOVA. Statistical analysis was undertaken using STATA (version 14). Concordance in delineation of volumes between observers were assessed using Dice Similarity Coefficient (DSC). Results GTVp and GTVn contours from MRI_1 and MRI_2 were assessed in ten patients; resulting in 20 MRI scans and 49 lesions delineated on T1PCFS sequences. There were no statistically significant differences in mean volume for GTVp (p=0.31) or GTVn (p=0.98) delineated on the T1PCFS sequences between observers.

The DSC between the RTT and CO on the T1PCFS was 0.66 (SD 0.16) for GTVp and 0.81 (SD 0.12) for GTVn. The RTT DSC with the CO_RO volumes was 0.64 (SD 0.25) and 0.83 (SD 0.14) for GTVp and GTVn, respectively.

No significant difference in measured mean ADC or % change in ADC was detected between observers, table 1.