Purpose:

Current positioning verification in radiotherapy treatment mainly relies on matching bony structures using X-ray based imaging techniques such as electronic portal imaging, cone beam CT and MV CT, where visualization of tumour and soft organ-at-risk (OAR) is relatively poor. With the development of advanced radiotherapy techniques, higher accuracy in positioning verification by imaging is demanded [1]. MR simulator (MR-sim) has the advantage of better soft tissue contrast and is gaining increasing interests in radiotherapy [2]. This study aimed to assess the positional repeatability of soft tissue and bony structure in head and neck on healthy volunteers using a commercial 1.5T MR-sim. By investigating the correlation between the positional deviations of various head and neck tissues from multiple scans, its implication to current clinical practice in radiotherapy positioning verification is discussed.

Method:

Four healthy volunteers were scanned (4 times for each subject) over a 2-month period using a T2w 3D-CUBE sequence (TR/TE=2000/130ms; voxel-size=0.8x0.8x1mm³; 3D geometric corrected) with a 1.5T dedicated MR-sim. Each volunteer was immobilized using a personally customized 5-point open-face thermoplastic mask and a standard head-rest. Permanent marks were labeled on thermoplastic mask as positioning reference. Volume of interests (VOI) of intervertebral discs between C₂ and C₃ (C₂₃), C₆ and C₇ (C₆₇), brainstem, pituitary gland, eye balls (eyes), bi-lateral parotid glands (PGs), mandible and larynx were drawn by a physicist specialized in MRI. Three-dimensional volume centroid was calculated for each VOI. Inter-scan positional shift was calculated as the VOI centroid displacement compared to the first scan of each subject. Correlations between inter-scan positional shifts on each direction (LR, SI and AP) between pairs of structures were quantified by Pearson correlation coefficient r. Two-tailed p value was also calculated.

Results:

Inter-scan positional shifts was plotted, and statistically significant correlation between pairs of structures were denoted in Fig. 1. Among all significantly correlated structure pairs, only shifts of pairs of C₂₃ and brainstem, pituitary and eyes, mandible and PGs were correlated in all three directions. A global significant correlation between all pairs of structures was found in SI direction. C₂₃ exhibited significant correlation with all other structures in AP direction. Averaged shifts and their range from all scans were summarized in Tab. 1, mandible exhibited relatively larger shifts in all directions, while brainstem and pituitary shifted relatively small.

Discussions:

High correlations in all directions can be found between some neighboring structures (they are C₂₃ and brainstem, pituitary and eyes, mandible and PGs), indicating that positioning verification of treatment target sites using remote structural landmarks might be problematic in head and neck. Furthermore, our data suggested that anatomical structures located superficially (mandible, larynx, PGs and eyes) could present larger inter-fractional positional variability than those located deeply (C₂₃, brainstem and pituitary). These probably attributed to their greater freedom degree of mobility and/or deformation under immobilization. Different from CT-sim, the couch of MR-sim is only movable in SI direction for scan, which partially explained the smaller range of tissue shifts in AP and LR directions. On the other hand, high correlations and larger range of shifts were noted In SI direction, which might be attributed to the relatively low MR-sim couch motion reproducibility and/or inadequate immobilization performance of the open-face thermoplastic mask in this direction, but yet to be further investigated.

Conclusions:

To conclude, current positioning verification practice for radiotherapy treatment, which mainly relies on bony structure, may be inadequate and inefficient in head and neck, as the inter-fractional shift pattern of treatment target tissues and OARs could be significantly different from that of bony structure, particularly when they are located distantly. MR-sim may play a potential role in improving head and neck radiotherapy by enabling positional verification directly on target tissues rather than simply on bony structure in the future. As the inter-fractional shift pattern of target tissues and OARs could be significantly different from that of bony structure, particularly when they are located distantly from targets, the current positioning verification practice for radiotherapy treatment relying mainly on bony structure may be inadequate and inefficient in head and neck. MR-sim may play a potential role in improving head and neck radiotherapy by enabling positional verification directly on target tissues rather than simply on bony structure in the future.