ESTRO 38 Abstract book

S667 ESTRO 38

MRI parameter thresholds for early detection of subclinical cases of ORN. EP-1204 Quantitative signal intensity kinetics of normal tissues of the head and neck on the MR-Linac. F. Gregucci 1,2 , H. Bahig 1 , A.S. Mohamed 1 , R. He 3 , B.A. Elgohary 1 , H. Lee 4 , Y. Ding 4 , J. Wang 4 , H. Elhalawani 1 , G.S. Ibbott 4 , M. Chetvertkov 4 , G. Bosco 4 , S. Ikner 4 , S.P. Ng 4 , D.C. Fuller 1 1 University of Texas MD Anderson Cancer Center, Radiation Oncology, Houston, USA ; 2 University of Padova, Radiation Oncology, Padova, Italy ; 3 University of Texas Medical School, Diagnostic and Interventional Imaging, Hoston, USA ; 4 University of Texas MD Anderson Cancer Center, Radiation Physics, Houston, USA Purpose or Objective The MR-Linac, a hybrid device combining a magnetic resonance with a linear accelerator, holds the promise of online anatomic and functional imaging which would allow daily radiotherapy adaptation. In this study, we sought to quantitatively assess the repeatability and reproducibility of signal intensity (SI) over time from serial T1- and T2- weighted magnetic resonance imaging (MRI) acquired on the MR-Linac with standardized acquisition parameters and under stable conditions. Material and Methods 6 healthy volunteers underwent 7 serial, immediately consecutive T1- and T2-weighted MRI studies of the head and neck (H&N) region using a 3 Tesla MR-Linac. Normal tissue volume of interest (VOI) were defined based on spin-echo sequences T1- and T2-weighted imaging. For a total of 42 MRI studies, we performed manual segmentation using T2-spin-echo sequence, for the following VOI: mandible, parotid glands, submandibular glands, sternocleidomastoid muscles, spinal cord and vertebral body. Deformable image registration was performed between the different MRI time points in T1 and T2 sequences, all VOI were then propagated and reviewed by expert radiation oncologists. From these serial MRI studies, a collection data concerning the VOI has been created bringing out a quantitative analysis of the following SI parameters: mean, max and min values. For the statistical analysis the VOI were grouped in the following 4 tissue categories: gland (G), bone (B), muscle (M) and nerve (N). Results The mixed model estimates on T1-MRI showed that max, min and mean values of SI were for G group -1002.88 (t=0.0014), -161 (t=0.0005) and -394.56 (t<0.0001), respectively; for B group -384.73 (t=0.178), -202.41 (t<0.0001) and -442.24 (t<0.0001), respectively; for M group -1008.08 (t=0.0013), 26.69 (t=0.513) and -91.18 (t=0.283), respectively; for N group 1962.63 (t=0.0004), 230.41 (t=0.0029) and 603.68 (t=0.0003), respectively. The mixed model estimates on T2-MRI showed that max, min and mean values of SI were for G group -9.97 (t=0.03), 11.39 (t<0.0001) and 6.49 (t=0.04), respectively; for B group -0.59 (t=0.89), 3.43 (t=0.12) and 12.44 (t=0.0008), respectively; for N group 66.75 (t<0.0001), 17.00 (t<0.0001) and 27.35 (t<0.0001), respectively. The F-test in T1- and T2-weighted MRI showed differences between the different VOI groups in max, min and mean SI values (f<.0001 for all) but no SI differences inside the same VOI group and between the different time points for each VOI groups. Conclusion The SI kinetics for analyzed tissue categories on T1 and T2-spin-echo sequences are reproducible and repeatable between different subjects and over the time, keeping unchanged the MRI acquisition parameters. These results could be an helpful tool for future studies of correlation between SI kinetics deviation and observed toxicity profile for the purpose of developing predictive models in H&N cancer.

specific informed consent. Eligibility criteria included; age>18 years, pathological evidence of head and neck malignancy with a history of curative-intent external beam radiotherapy; patients with clinically confirmed high-grade ORN requiring surgical intervention; and no contraindications to MRI. Prior to DCE-MRI, T1 mapping was performed using a total of 6 variable flip angles. The DCE-MRI acquisition consisted of a 3D SPGR sequence to gain a sufficient signal-to-noise ratio (SNR), contrast, and temporal resolution. Images were acquired every 5.5 s for a total of 5 minutes. Extended Toft’s pharmacokinetic model was used for analysis. Motion correction was applied to enhance the quality of map computation. Manual segmentation of advanced ORN 3-D volume was done using anatomical sequences (T1, T2, and T1+contrast) to created ORN volumes of interest (ORN- VOIs). Subsequently, normal mandibular VOIs were segmented on contralateral healthy mandible of similar volume and anatomical location (i.e. mirror image) to create self-control VOIs. Finally, anatomical sequences were co-registered to DCE sequences and contours were propagated to respective Ktrans and Ve quantitative parameter maps. The workflow is summarized in Figure 1.

Results Seventeen patients with advanced ORN requiring surgical resection of the mandible were analyzed. Median age at diagnosis was 59 years (range 38-72), and 88% were men. Primary tumors were localized to the oropharynx (n=9), oral cavity (n=4), salivary glands (n=2) and nasopharynx (n=2). All patients were treated with intensity modulated radiotherapy (IMRT) to a median dose of 70 Gy in 33 fractions. Using matched pairs analysis, we measured higher Ktrans and Ve values in ORN-VOIs compared with control regions (0.49 vs 0.19 min −1 , p<0.0001 and 1.2 vs 0.6, p=0.01; respectively). The median relative increase of Ktrans in ORN-VOIs was 2.7 fold greater than that of healthy mandibular control VOIs (range 1.4-7.8); the relative increase of Ve in ORN-VOIs was 2.6 fold greater than that of healthy mandibular control VOIs (range 1.1- 8). Figure 2 depicts boxplots of ORN vs Control VOIs parameters.

Conclusion Our results confirm there is a quantitively significant higher degree of leakiness in the mandibular vasculature as measured using DCE-MRI parameters of areas affected with advanced grade of ORN versus healthy mandible. We were able to measure significant increases in both parameters (2.7 fold Ktrans, 2.6 fold Ve) compared to values from non-ORN mandibular bone. Further efforts are ongoing to validate these findings and to establish DCE-