Imaging, Registration & Immobilization

1.The CT-MRI partnership, acquisition quality, and image age

A thin, contiguous planning CT is the geometrically faithful reference and supplies electron density for dose calculation, while MRI supplies the soft-tissue definition for most cranial targets. The two are rigidly co-registered. The planning MRI should use a stereotactic protocol with an isotropic or near-isotropic dataset, no interslice gap, and distortion correction appropriate to the scanner and sequence. For brain metastases, the planning post-contrast 3D T1 dataset should be ≤1 mm: thicker images can miss small lesions and materially change their contoured volume.

Image quality is not enough if the anatomy has changed. Planning imaging should be obtained as close to treatment as practical. For brain metastases, a useful operational standard is MRI acquired within 7 days of SRS; same-day stereotactic imaging remains ideal for frame-based Gamma Knife workflows. Repeat or update imaging sooner when the target is large or rapidly changing, when edema or corticosteroid dosing has shifted, after recent surgery, or when the neurologic examination changes. A diagnostic MRI several weeks old may establish the indication, but it should not automatically define the treatment volume.

2.Indication-specific MRI and adjunct imaging

The core tumor sequence is best described without tying it to one vendor: a thin-slice isotropic 3D post-contrast T1-weighted acquisition, commonly an inversion-prepared gradient-echo sequence such as MPRAGE, SPGR/BRAVO, or TFE. For metastases, post-contrast 3D turbo spin echo or black-blood techniques such as SPACE, CUBE, or VISTA can complement the gradient-echo dataset by suppressing vascular signal and improving conspicuity of small lesions. Specialized targets need specialized anatomy:

FGATIR and tractography add anatomical context for Vim targeting but should not be presented as infallible direct visualization. The thalamic target remains a synthesis of stereotactic landmarks, atlas anatomy, patient-specific imaging, and the treatment team's validated method. Likewise, a CT-only cranial workflow is a fallback for MRI contraindication, not an equivalent default; the loss of soft-tissue discrimination must be recognized explicitly.

3.MRI geometric distortion: the quiet hazard

MRI is geometrically imperfect. Gradient nonlinearity, main-field inhomogeneity, and susceptibility and chemical-shift effects can displace anatomy by clinically meaningful distances, worsening toward the periphery of the field of view and at air/bone interfaces (skull base). In one radiosurgery series the median target displacement caused by gradient nonlinearity alone was 1.2 mm with a maximum of 3.9 mm, and 8 of 28 cases met a predefined criterion for geometric miss; AAPM Task Group 284 reports residual distortion after vendor correction exceeding 1 mm beyond roughly 10 cm from magnet isocenter. Because radiosurgery uses essentially no margin, an uncorrected distortion places the contour, and the dose, off target. Mitigation includes vendor distortion-correction algorithms, which correct system gradient nonlinearity but not patient-induced susceptibility distortion, distortion-optimized sequences with high receiver bandwidth, deliberate phase-encode direction choice, centering the head at magnet isocenter rather than at the coil, QA phantoms to characterize residual distortion at commissioning and at least annually, and, critically, using the geometrically faithful CT as the spatial reference and verifying the fusion on multiple planes and landmarks rather than trusting an automated result.

4.Frame, mask, and body systems

Immobilization trades invasiveness for repeatability:

• Invasive stereotactic frame (e.g., Leksell coordinate frame): pinned to the skull, it provides a rigid coordinate system and the highest single-session accuracy (sub-millimeter), the historical standard for single-fraction cranial SRS.
• Thermoplastic mask (frameless): non-invasive and necessary for most fractionated workflows. Validated image-guided systems can achieve sub-millimeter end-to-end accuracy, but masks show greater setup and intrafraction variability than rigid frames; performance belongs to the complete chain of mask, image guidance, correction, motion threshold, and beam-hold policy, not to the mask alone.
• Body immobilization: vacuum bags and stereotactic body frames for spine/trunk, with abdominal compression, breath-hold, or gating to manage respiratory motion for thoracic/abdominal SBRT.

The key point is that frameless accuracy is not a property of the mask alone: it depends on the image-guidance and motion-monitoring system that accompanies it.

5.Seeing and correcting position before and during delivery

Modern delivery is image-guided. Cone-beam CT verifies volumetric position; stereoscopic kV imaging (e.g., ExacTrac) localizes bony anatomy and supports six-degree-of-freedom (6DOF) couch correction; optical surface monitoring (e.g., surface-guided systems) tracks the patient surface for setup and intrafraction motion, triggering a beam hold if the patient moves. Spine SBRT relies on bony (vertebral) tracking; mobile body targets use implanted fiducials and respiratory tracking (e.g., the robotic Synchrony approach). The radiation-isocenter accuracy underlying all of this is verified by the Winston-Lutz test and the broader QA program on the planning and QA page.