Synopsis

Keywords: Cancer, Head & Neck/ENT, Tumour Hypoxia, Radiotherapy, OE-MRI, MR Linac

Hypoxia is common to most solid tumours and mediates treatment resistance. Oxygen-enhanced (OE)-MRI can identify, map and quantify hypoxia in multiple tumour types including H&N cancer. We build upon previous work to show that OE-MRI biomarkers: (i) can be deployed on multiple MRI systems; (ii) are repeatable; (iii) can detect cohort level response to radiotherapy; (iv) can identify which individual lesions have significant changes in their oxygenation, hypoxic volume and hypoxic fraction.

Introduction

Hypoxia is seen in solid tumours, and its presence and extent, indicates poor prognosis and treatment resistance. Oxygen enhanced (OE)-MRI shows promise as a non-invasive method of quantifying hypoxia¹. Our previous work shows that OE-MRI can detect hypoxia modification², is feasible in the head and neck (H&N) region and is translatable to MR-guided radiotherapy systems³. Both the oxygen-induced change in longitudinal relaxation rate (ΔR₁) and the part of the tumour that is perfused but insensitive to oxygen change (oxygen refractory (pOxy-R)) have been used as biomarkers^1,2,3,4. Here, we sought to assess radiotherapy-induced treatment effects in patients with H&N cancer at both a cohort and individual level on two MR systems.

Methods

Patients with H&N squamous cell cancer were recruited after research ethics approval and written informed consent. Imaging was performed on either a 1.5 T Philips Ingenia MR-RT (diagnostic MR) or a 1.5 T Elekta-Philips Marlin (MR linac) system. Each underwent (chemo)radiotherapy.

Patients were imaged on the diagnostic MR (N = 14) and MR linac (N = 5) systems. Imaging included T₁ measurement (3D inversion recovery turbo field echo (IRTFE), 3x3x5mm, TI = 100, 500, 800, 1100, 4300ms) and dynamic OE acquisition using the same IRTFE sequence with TI = 1100 ms, temporal resolution = 12s. Oxygen was delivered via a gas blender at a flow rate of 15 l/min and high concentration oxygen mask, alternating between medical air (dynamics 1-25), 100% oxygen (dynamics 26-70), and medical air (dynamics 71-91). Dynamic contrast enhanced (DCE)-MRI was acquired using a T₁-FFE Dixon sequence (3x3x5mm, 3.8s temporal resolution) with IV contrast agent injection (Dotarem, 3 ml/s, 0.2ml/kg). A post contrast T₁-weighted acquisition with fat saturation permitted lesion volume delineation. A Eurospin TO5 phantom was used to evaluate native T₁ measurement repeatability and drift.

Image processing and analysis was carried out in MATLAB (Mathworks). Motion correction was carried out using Elastix⁵. From 19 patients, a total of 32 lesions including primary tumour and involved lymph node volumes were delineated by a H&N oncologist. T₁ maps permitted conversion of signal to ΔR₁ (= R_1,O2 - R_1,air) during oxygen inhalation. DCE-MRI signal and dynamic OE-MRI signal data were combined to identify voxels which were significantly enhancing (p<0.05) between pre- and post-contrast or air and 100% oxygen. Voxels which were identified as enhancing on DCE-MRI but not on OE-MRI data were classed pOxy-R, indicating hypoxia. The pOxy-R volume was also measured. The repeatability coefficient (RC) and within-subject coefficient of variation (wCV) were estimated for each parameter from patients who underwent both pre-treatment scans (N = 12). Paired t-tests assessed change in derived parameters compared to baseline measurements.

Results

T₁ measurements acquired in the phantom on both diagnostic and MR linac systems were highly repeatable (Figure 1). Difference in mean T₁ value on the diagnostic MRI was 4 (±6) ms (RC = 14 ms, wCV = 1%). In comparison, difference in mean T₁ value on the MR Linac was 7 (±7) ms (RC = 18 ms, wCV = 1%). Comparison of the two MRI systems showed difference in mean T₁ of 8 (±18) ms (RC = 37 ms, wCV = 2%). ΔR₁ time-courses for gels in the Eurospin phantom indicated negligible signal drift on both systems.

All imaging was well tolerated. Lesion measurement repeatability (across the two systems) was: RC = 0.016 s^-1, wCV = 29% for ΔR₁, RC = 0.16, wCV = 15% for pOxy-R fraction, and RC = 7cm³, wCV = 13% for pOxy-R volume (figure 2). A significant increase in ΔR₁ was found at the cohort level at both weeks 2 (p < 0.001) and 4 (p = 0.006) of treatment. Significant decreases in pOxy-R fraction were found at weeks 2 (p < 0.001) and 4 (p = 0.004) of treatment, accompanied by significant decreases in pOxy-R volume at weeks 2 (p = 0.002) and 4 (p = 0.03) of treatment (figure 3). The number of lesions exhibiting change greater than measurement RC was determined for each parameter (figure 4; data shown for week 2). Reduction in hypoxia (within 95% CI of RC) was seen in: 11/26 for ΔR₁, 15/26 for pOxy-R fraction, 11/26 for pOxy-R volume. 8/26 of whole lesion volume (WLV) reduced, of which 4/8 WLV reductions were independent of hypoxia change. In one case hypoxia was increased following therapy (pOxy-R fraction 1/26).

Discussion and Conclusion

We investigated OE-MRI assessment of changes in lesion hypoxia in patients with H&N cancer following radiotherapy. We show:
1. OE-MRI biomarkers can be deployed across multiple MRI systems at 1.5 T.
2. Good repeatability of OE-MRI derived parameters: ΔR₁, pOxy-R fraction and pOxy-R volume.
3. Significant cohort level changes in these parameters from pre-treatment to 2 weeks of (chemo)radiotherapy, that were maintained at 4 weeks.
4. Repeatability assessment enables identification of which lesions had significant change in hypoxia.
5. Hypoxic changes were independent of WLV changes.

Overall, we conclude that OE-MRI can monitor changes in lesion hypoxia during therapy^1,2. Both the volume of hypoxic lesion and the fraction of hypoxia within the lesion were quantified and may be putative biomarkers for future multi-centre assessment of patients with hypoxic lesions on MR linac systems.

Acknowledgements

This work was supported by funding from the Medical Research Council and Cancer Research UK.