Frame vs. frameless cranial fixation and the setup-accuracy budget behind PTV margins
Stereotactic radiosurgery was born with the invasive rigid frame. Devices like the Leksell frame (Gamma Knife) and the Brown-Roberts-Wells (BRW) / Cosman-Roberts-Wells (CRW) frames are bolted directly to the outer table of the skull with four titanium or carbon-fiber pins placed under local anesthesia, creating a fixed mechanical reference between the patient's bony anatomy and the treatment coordinate system.
The rigid frame works because bone, unlike skin or muscle, does not deform, slide, or relax over the course of a treatment session. Once four pins are seated into the outer table of the skull — typically two frontal and two occipital/parietal — the frame ring becomes a fixed extension of the skull itself. A stereotactic localizer box is then attached to the frame, and imaging (CT/MRI) performed with the localizer in place allows every voxel of the images to be mapped to an (x,y,z) frame coordinate with sub-millimeter fidelity.
Because the frame-to-target relationship cannot change once the pins are seated, there is essentially no immobilization uncertainty left in the accuracy budget — the dominant remaining error sources are imaging distortion, frame manufacturing tolerance, and the mechanical accuracy of the treatment unit itself. This is why frame-based Gamma Knife and linac radiosurgery have historically reported total system accuracies of roughly 0.1–0.3 mm.
Because the frame is a single-use, single-day device, historically frame-based SRS was almost always delivered as one large fraction — fractionation (splitting the dose over several days) was mechanically impractical, since re-pinning the frame each day is neither comfortable nor perfectly reproducible.
The rigid frame's accuracy comes at a real cost to the patient and workflow:
• Invasiveness: skull pins require local anesthetic infiltration, carry a small risk of bleeding, infection, or pin-site discomfort, and are generally not tolerated by anxious or pediatric patients without sedation • Same-day workflow: imaging, planning, and treatment must all occur while the frame remains attached — this compresses the entire process into a single long day and removes the option of fractionation, which limits its use for tumors that benefit radiobiologically from dose splitting (e.g., larger lesions near critical structures) • Single target session: the frame cannot be removed and precisely reattached with the same accuracy, so any need to re-treat requires a completely new procedure • Physician/staff burden: pin placement requires a proceduralist and adds real procedural time and resource cost compared to a mask fitting
These trade-offs are exactly what motivated the development of non-invasive, frameless immobilization systems — provided that image guidance could compensate for the accuracy lost by giving up rigid bone fixation.
Despite the shift toward frameless workflows, rigid frames remain in routine use, particularly for Gamma Knife radiosurgery of small, well-defined intracranial targets (arteriovenous malformations, small metastases, functional targets such as trigeminal neuralgia or essential tremor) where single-fraction treatment is clinically appropriate and the frame's intrinsic accuracy is fully exploited. Newer "frame-based-but-relocatable" hybrid systems have also been developed to try to retain rigid-fixation-level accuracy while allowing multi-day fractionation, though they have not fully displaced either the classic pinned frame or fully frameless mask-based platforms.
The frameless era of cranial radiosurgery replaced skull pins with a custom-molded thermoplastic mesh mask, heated and draped over the patient's face and head, then clipped rigidly to the treatment couch or a fixed baseplate. This trades a small amount of positional accuracy for a large gain in patient comfort, safety, and the ability to fractionate treatment across multiple days.
A sheet of perforated thermoplastic is heated in a water bath (~65–70°C) until pliable, then draped over the patient's face, forehead, and chin while still warm and molded to the individual's contours. As it cools over a few minutes it hardens into a rigid, patient-specific shell that is then clipped at 3–5 points to a rigid baseplate fixed to the couch or headrest. Modern high-precision SRS masks (e.g., 5- or 6-point systems, or open-face designs that leave the eyes and nose exposed for optical tracking) are specifically engineered for tighter tolerances than standard palliative-radiotherapy masks.
Unlike a frame, the mask relies on soft-tissue and skin contact rather than direct bone fixation. Skin, subcutaneous fat, and small amounts of mask flex all introduce potential slack — so the mask's effective rigidity is only as good as how snugly and reproducibly it compresses the underlying soft tissue against the (comparatively fixed) skull.
A well-fitted 5-point stereotactic mask system, combined with image guidance, can approach — but typically does not fully match — the sub-millimeter accuracy of a pinned frame; published series generally report residual setup uncertainties in the 0.5–1.5 mm range without intrafraction correction.
The single most important clinical consequence of the mask system is reusability: the same mask can be refitted to the same patient, day after day, with good (though not perfect) reproducibility. This makes fractionated stereotactic radiotherapy (SRT) — delivering the stereotactic dose over 3–5 sessions instead of one — practical for the first time.
Fractionation is radiobiologically valuable for larger lesions or targets close to critical normal structures (optic apparatus, brainstem), because splitting dose over multiple sessions allows normal tissue to repair sublethal damage between fractions while tumor cells accumulate damage — improving the therapeutic ratio compared to a single large fraction. This benefit was essentially unavailable with a single-use pinned frame.
Because a mask does not fix directly to bone, several factors can introduce small day-to-day and intrafraction positional errors:
• Soft-tissue compression variability: how tightly the mask is clipped can vary slightly between fractions and even relax slightly during a session • Mandible motion: jaw movement (swallowing, talking, clenching) can shift the lower face without moving the cranium, which is why bite-blocks are often added for extra stability • Mask fatigue: repeated heating/cooling or repeated donning across fractions can slightly change the mask's fit over a treatment course • Patient factors: weight change, swelling, or anxiety-related muscle tension over a multi-week course
These are precisely the uncertainties that pre-treatment and intrafraction image guidance are designed to detect and correct.
A custom dental bite-block — a mouthpiece molded from a dental impression of the patient's upper teeth — offers a clever way to recover some of the rigidity lost by moving away from a bone-pinned frame, without any invasive procedure. It exploits the fact that the maxilla (upper jaw) is a fixed part of the skull base, unlike the independently mobile mandible.
The skull is not one uniformly rigid unit from a fixation standpoint — the mandible (lower jaw) articulates at the temporomandibular joint and can move independently through opening, closing, or lateral drift, while the maxilla is sutured to the base of the skull and, for practical stereotactic purposes, moves as a single rigid body with the cranium. A bite-block that indexes precisely against the upper teeth therefore inherits much of the same rigidity as a bone-referenced fixation point, without requiring any skin-penetrating hardware.
A custom bite-block is fabricated from a dental alginate or thermoplastic impression of the patient's upper dental arch, producing a mouthpiece that can only seat in one reproducible position — analogous to a key that fits only one lock. When the patient bites down onto the block and it is rigidly linked to the immobilization frame or mask baseplate, upper-jaw (and therefore cranial) position becomes far more tightly constrained than mask contact with facial skin alone.
Bite-blocks are frequently combined with a thermoplastic mask rather than used alone: the mask constrains gross head position and rotation, while the bite-block removes the residual slack from jaw drift and soft-tissue compression — together approaching frame-like reproducibility without any pins.
A dental technologist or radiation therapist takes an impression of the patient's upper teeth (occasionally both arches) at the simulation appointment. The resulting mouthpiece is rigidly mounted to the immobilization baseplate, headrest, or mask frame, positioned so the patient bites into the same indexed position at every session. Because the impression captures the fine detail of individual tooth surfaces, the bite is essentially non-reproducible in any other orientation — providing a built-in check that the patient is correctly seated before imaging even begins.
Limitations exist: patients who are edentulous (no teeth), have significant dental work changes between simulation and treatment, or cannot comfortably hold a fixed bite for the treatment duration are not good candidates. For these patients, mask-only or frame-based fixation remains necessary.
In terms of achievable setup accuracy, bite-block-augmented systems generally fall between a standalone thermoplastic mask and a pinned rigid frame — commonly cited residual uncertainties in the 0.3–0.8 mm range when properly combined with a mask and verified with image guidance. This makes bite-block combinations an attractive middle ground for centers wanting to deliver fractionated, image-guided cranial SRS/SRT with accuracy approaching frame-based systems, but without ever placing skull pins.
No immobilization device — frame, mask, or bite-block — is trusted blindly at the moment of treatment. Modern cranial SRS/SRT workflows verify actual patient position against the planned reference position using CBCT, orthogonal kV imaging, or optical surface tracking immediately before (and increasingly throughout) each fraction, and apply a measured couch correction before the beam is turned on.
Immediately before beam-on, a verification image (cone-beam CT, orthogonal kV pairs against digitally reconstructed radiographs, or a surface scan matched to the planning CT surface) is acquired with the patient immobilized in treatment position. Automated or semi-automated rigid registration compares this image to the planning reference, producing a measured translational shift (x, y, z) and, on six-degrees-of-freedom couches, rotational corrections (pitch, roll, yaw) as well.
If the measured shift exceeds a pre-defined tolerance (commonly on the order of 0.5–1 mm / 0.5–1° for cranial SRS), the treatment couch is automatically or manually adjusted to bring the patient into alignment, and — for high-precision protocols — a repeat image confirms the correction before treatment proceeds.
Image guidance cannot create accuracy that isn't there — it can only detect and correct the shift that has already occurred. Its value is proportional to how large and how variable the underlying immobilization uncertainty is: a loosely fitting mask benefits far more from IGRT than an already-rigid pinned frame.
Two distinct verification strategies are used clinically, and they compensate for different failure modes:
• Pre-treatment-only verification: a single CBCT or kV pair is acquired and corrected just before beam-on. This catches setup error at the start of the session but cannot detect drift that occurs during a multi-minute treatment delivery (e.g., patient relaxing into the mask, small head movement, coughing) • Continuous intrafraction monitoring: optical surface-tracking systems (structured light or stereoscopic camera arrays) or periodic intra-beam imaging track patient position throughout delivery, automatically pausing the beam if motion exceeds tolerance
Because frameless mask-based patients are more prone to slow intrafraction drift than pinned-frame patients, continuous monitoring is especially valuable for frameless SRS/SRT — it can recover much of the accuracy gap that giving up rigid bone fixation otherwise creates.
Image guidance itself is not error-free: image resolution, imaging artifact, soft-tissue deformation between planning and treatment, and the registration algorithm's intrinsic precision all contribute a small residual uncertainty — typically a fraction of a millimeter for high-quality stereotactic CBCT registration. This registration/imaging uncertainty must be added (in quadrature, since the error sources are independent and random) to the residual immobilization uncertainty and the mechanical accuracy of the delivery system itself to arrive at the total geometric uncertainty that determines the safe PTV margin — the subject of the final stage.
Every source of geometric uncertainty in the SRS chain — immobilization, image-guidance registration, and mechanical delivery accuracy of the linac or robotic arm — combines to determine the smallest safe margin that can be placed around the visible tumor (GTV/CTV) to guarantee full dose coverage. Because SRS derives its clinical advantage from an extremely steep dose fall-off just outside the target, the size of this margin has an outsized effect on both efficacy and toxicity.
Because immobilization error, imaging/registration error, and machine mechanical error arise from largely independent physical processes, they are combined not by simple addition but in quadrature (root-sum-square):
σ_total = √(σ_immob² + σ_imaging² + σ_mechanical²)
This is the standard statistical treatment for combining independent random uncertainties, and it means that reducing the single largest contributor yields the greatest overall benefit — which is exactly why frameless programs invest heavily in image guidance to compensate for giving up rigid bone fixation: shrinking the immobilization term (σ_immob) via IGRT correction lowers σ_total even though σ_imaging and σ_mechanical stay roughly fixed.
The resulting σ_total (often reported as a 1-sigma or a 95% confidence "van Herk"-style margin recipe) sets the minimum PTV margin required to ensure the CTV receives full prescription dose despite realistic day-to-day and intrafraction positioning error.
Van Herk's widely used margin recipe (developed for conventional fractionated radiotherapy but conceptually applied to SRS accuracy budgets) combines systematic and random errors with different weightings, since systematic (constant, day-to-day) errors shift the whole dose distribution while random (fraction-to-fraction) errors blur its edges — both must be accounted for, not just averaged.
SRS/SRT is specifically valued for its steep dose gradient: prescription-level dose falls off to a small fraction of that dose within just a few millimeters outside the target, sparing adjacent critical structures (brainstem, optic chiasm, cochlea, normal brain) that a conventional radiotherapy field could not spare. Adding PTV margin directly works against this advantage in two ways:
• It increases the physically treated volume — for a small (~1 cm diameter) target, expanding the margin from 1 mm to 3 mm roughly quadruples the treated volume, meaningfully increasing the dose delivered to surrounding normal brain and the risk of radionecrosis • It blurs the sharp fall-off itself — a larger margin effectively pushes the high-dose gradient further from the true tumor edge, reducing the separation between prescription isodose and nearby organs at risk
This is why, for SRS specifically (far more than for conventional fractionated radiotherapy of larger targets), minimizing the accuracy budget — and therefore the PTV margin — has a direct, outsized effect on the clinical safety margin available around critical intracranial structures.
Published comparisons generally support the following practical picture: rigid pinned frames combined with careful stereotactic imaging can support near-zero (roughly 0–1 mm) PTV margins, reflecting their historically demonstrated ~0.1–0.3 mm mechanical accuracy. Frameless mask-based systems without rigorous image guidance may require margins toward the higher end of the 1–3 mm range to safely account for their larger and more variable setup uncertainty. However, when frameless systems are paired with rigorous image guidance — pre-treatment CBCT verification and, ideally, continuous intrafraction optical or imaging monitoring — many modern series report total accuracy approaching that of frame-based systems, allowing margins in the 1–2 mm range even without any invasive fixation.
The clinical decision, then, is not simply "frame versus frameless" but a full system-level choice: immobilization hardware, image-guidance protocol, and delivery-system mechanical accuracy must be evaluated together against the margin they can jointly support for a given target and its proximity to critical structures.