Radiation therapy is a cornerstone in the management of head and neck squamous cell carcinoma, yet patient outcomes and treatment-related toxicity vary considerably. This variability has driven interest in quantitative imaging markers that could support earlier and more individualised treatment decisions. Diffusion-weighted MRI and the apparent diffusion coefficient provide a non-invasive way to characterise tumour tissue and monitor biological change during therapy. Hybrid MRI-guided linear accelerator systems allow repeated diffusion imaging in the treatment position, offering new opportunities for adaptive radiotherapy. However, these opportunities depend on how reliably apparent diffusion coefficient (ADC) can be measured over time. Before ADC changes can be used to guide clinical decisions, it is essential to understand the level of measurement variability inherent to MR-Linac diffusion imaging.
Technical Constraints of Diffusion Imaging on MR-Linac
MR-Linac systems integrate a 1.5-T MRI scanner with a linear accelerator, enabling imaging and radiation delivery within the same platform. While this configuration supports longitudinal imaging, it introduces technical differences compared with diagnostic MRI systems. The split-gradient coil design required for radiation beam passage can lead to gradient non-linearities. Lower gradient strength and slew rate can reduce signal-to-noise ratio by requiring longer diffusion times to achieve the same b values. In addition, the relatively small number of receive channels in the system’s body coil further limits signal quality.
These factors directly affect diffusion-weighted imaging and the precision of ADC measurements. Even so, previous work has shown that optimised echo-planar imaging diffusion sequences on MR-Linac systems can achieve repeatability and reproducibility comparable to diagnostic scanners, although absolute ADC values may differ across platforms. For treatment monitoring, internal consistency on a single system is more important than cross-platform agreement. Test–retest imaging provides a practical method to define the threshold beyond which ADC changes are unlikely to be explained by measurement noise alone.
Test–Retest Performance in Tumours and Lymph Nodes
ADC reproducibility was evaluated in patients with head and neck squamous cell carcinoma who underwent two diffusion MRI examinations on a 1.5-T MR-Linac. Imaging was performed before the start of radiation therapy or, in a small subset, immediately before the second or third treatment fraction. Primary tumours and pathological lymph nodes were analysed separately, reflecting typical clinical disease patterns.
Lesions were contoured on b0 images with reference to T2-weighted imaging, and mean and median ADC values were calculated across the full tumour volumes. Reproducibility was quantified using established metrics, including the reproducibility coefficient and the percent reproducibility coefficient. These values represent the minimum absolute or relative ADC change that can be interpreted as a real change with high confidence.
Across the cohort, the reproducibility coefficients for mean ADC were similar for primary tumours and lymph nodes, corresponding to percent reproducibility coefficients of approximately 30%. Median ADC values showed comparable behaviour, with slightly higher variability that was not considered clinically meaningful. There was no clear evidence that scan interval affected reproducibility within the observed range, and lesion volume showed only weak associations with variability. Smaller lesions tended to show higher variability, although this trend did not reach statistical significance.
Must Read: Gadolinium Use Remains Central in Head and Neck MRI
Importantly, a subgroup analysis including patients who had received one or two radiation fractions before the second scan did not show worse reproducibility. In fact, reproducibility metrics were slightly better in this group, suggesting that early treatment exposure did not inflate measurement variability.
Implications for Adaptive Radiotherapy
ADC changes during radiation therapy have been associated with treatment response in head and neck cancer, but reported thresholds vary widely across studies. On the MR-Linac configuration assessed, ADC changes smaller than roughly 30% may fall within normal measurement variability. As a result, subtle biological changes that could still be clinically relevant may not be detectable with confidence.
Larger ADC changes, however, exceed the reproducibility threshold and remain detectable. These larger shifts may still be useful for identifying marked treatment response or resistance, depending on clinical intent. Mean and median ADC performed similarly, suggesting flexibility in metric selection without major impact on reliability. The findings also highlight practical influences on reproducibility, including susceptibility to geometric distortion in echo-planar imaging near air–tissue interfaces and increased variability in smaller lesions.
Test–retest evaluation of diffusion-weighted imaging on a 1.5-T MR-Linac shows that ADC measurements in head and neck squamous cell carcinoma are reproducible at a level that supports detection of large diffusion changes in both primary tumours and lymph nodes. At the same time, the reproducibility threshold limits sensitivity to smaller ADC shifts that may occur early during treatment. These results define a realistic baseline for interpreting ADC changes in MR-Linac–based adaptive radiotherapy and underscore the need for continued technical optimisation to improve sensitivity.
Source: Radiology: Imaging Cancer
Image Credit: iStock
