Multi-isocenter Cobalt-60 convergent beam dose planning for stereotactic radiosurgery
Gamma Knife radiosurgery, invented by Swedish neurosurgeon Lars Leksell in 1968, rests on a single mechanical guarantee: the target and the treatment machine share one fixed, unmoving coordinate frame. Before any beam is planned, a rigid stereotactic frame is bolted to the patient's skull and the lesion is localized within that frame's coordinate system to sub-millimeter precision.
The Leksell stereotactic frame is a rigid aluminum or carbon-fiber ring, fixed directly to the outer table of the skull with four pins under local anesthesia. Because bone does not deform, the frame creates an invariant Cartesian (x, y, z) coordinate space that is identical whether measured on the imaging table or the treatment couch. Every millimeter of the frame's geometry is known, so any point localized on an MRI, CT, or angiogram acquired with the frame attached can be translated directly into treatment coordinates without any additional registration step.
This is the conceptual core of stereotactic radiosurgery: instead of tracking the patient's anatomy in real time, the anatomy is mechanically pinned to a known reference frame, and the treatment unit is built to deliver dose to fixed points within that frame with extraordinary reproducibility.
Frame-based fixation achieves mechanical targeting accuracy of roughly 0.1–0.3 mm — a precision that has changed little since Leksell's original design because the physics of a rigid bone-anchored frame is essentially unimprovable.
Once the frame is attached, the patient undergoes high-resolution imaging — typically thin-slice T1 MRI with contrast, often fused with CT for bone detail and, for vascular lesions, digital subtraction angiography (DSA). Each modality has a fiducial box or N-localizer attached to the frame during the scan, allowing the treatment planning software (Leksell GammaPlan) to compute exact frame coordinates for every voxel.
Multimodal fusion matters because different lesions declare themselves differently: an arteriovenous malformation (AVM) nidus is best delineated on angiography, an acoustic neuroma or metastasis on contrast MRI, and critical structures like the optic apparatus, brainstem, and cochlea are best contoured on high-resolution MRI. The target volume and every organ-at-risk are contoured in this shared frame space before a single beam is planned.
The classical Leksell frame requires pin fixation on the day of treatment, limiting Gamma Knife to single-session (single-fraction) use. The modern Gamma Knife Icon adds a mask-based immobilization option with cone-beam CT verification and continuous infrared intrafraction motion monitoring, allowing fractionated treatment (multiple sessions) for larger lesions or those near critical structures, while still achieving sub-millimeter accuracy (<0.5 mm) through automatic beam gating if the patient moves beyond threshold.
This dual capability — pin-based rigid fixation for single-fraction radiosurgery, or mask-based relocatable fixation for hypofractionation — lets clinicians choose the immobilization strategy that best matches lesion size, location, and proximity to radiosensitive structures.
The defining hardware of Gamma Knife is its hemispheric array of 192 or 201 sealed Cobalt-60 sources, each emitting a narrow gamma-ray beam through a secondary and tertiary collimator. Mechanically, every one of these beams is aimed at exactly the same point in space — the isocenter — so that only where all beams overlap does the dose become therapeutic, while the entry path for each individual beam is exposed to negligible dose.
Each individual Co-60 beam is far too weak, and its path far too narrow, to injure tissue on its own — a single beam delivers only a small fraction of the prescription dose along its entire trajectory through the skull. Because 192–201 such beams are arranged on a hemisphere and all converge mechanically on one isocenter, the dose contributions add arithmetically only at the convergence point. Healthy brain along any single beam path sees dose spread across 192+ different trajectories, while the isocenter itself receives the full summed dose.
This convergent, non-coplanar beam geometry is what produces radiosurgery's hallmark dose distribution: a compact, roughly spherical high-dose volume surrounded by a dose gradient that falls away extremely quickly — because there is no single entry or exit path carrying a large dose burden.
The size of the spherical dose cloud produced at the isocenter is controlled by collimators. Earlier Gamma Knife models (the 4C and U) used interchangeable helmets pre-drilled with 4 mm, 8 mm, 14 mm, or 16 mm collimator channels, physically swapped between shots. The modern Perfexion and Icon platforms replace the helmet with a single fixed collimator body containing all three (4/8/16 mm) collimator sizes for every one of 8 independently movable sectors, so collimator size is now selected electronically per sector rather than by swapping hardware.
A 4 mm collimator produces a tight, small dose sphere suited to trigeminal nerve targets; a 16 mm collimator produces a broad sphere suited to covering large single lesions in fewer shots — with an inherent trade-off between shot count and conformality.
A single isocenter produces a dose cloud that is, to a first approximation, spherical or ellipsoidal. This is ideal for targets that are themselves small and roughly round — a small brain metastasis or an early acoustic neuroma. But most clinically relevant targets (AVM nidi, irregular metastases, tumors wrapping around a nerve) are not spherical. Forcing one large sphere to cover an irregular volume either under-doses the lesion's extremities (leaving a tail of tumor or nidus untreated) or over-doses adjacent normal brain to compensate — exactly the problem multi-isocenter planning was developed to solve.
Real lesions are rarely spherical. Leksell GammaPlan and its predecessors solve this by packing multiple isocenters — clinically called "shots" — at different positions within and around the target, so that the union of many overlapping small spheres approximates the true, irregular shape of the lesion far better than any single large sphere could.
Each isocenter placement, with its own (x,y,z) coordinate, collimator size, and relative weighting, is called a shot. A treatment plan is a collection of shots whose overlapping spherical dose clouds are summed together. Early Gamma Knife planning was entirely manual (forward planning): a physicist or neurosurgeon iteratively placed and resized shots by eye, checking the resulting isodose lines against the contoured target on sequential slices, refining positions until coverage and conformality were acceptable.
Classic targets that demand multiple shots include arteriovenous malformations (AVMs) with irregular, often serpentine nidi; vestibular schwannomas (acoustic neuromas) that curve along the internal auditory canal; and elongated or multi-lobulated brain metastases.
Modern GammaPlan versions offer inverse planning algorithms that work backward from clinical objectives — target coverage, selectivity (fraction of the prescription volume that is inside the target), and dose constraints on nearby organs at risk — to automatically propose an initial shot configuration, which the planner then refines. Inverse optimization can rapidly explore far more shot combinations than manual placement, often achieving better conformality with fewer total shots and shorter beam-on time.
Key plan-quality metrics include the Paddick conformity index (how tightly the prescription isodose volume matches the target volume), selectivity, and the gradient index — together quantifying how well dose is concentrated in the target and how quickly it falls away outside it.
Multi-isocenter logic even extends to functional (non-tumor) radiosurgery. For trigeminal neuralgia, a single small 4 mm shot is placed directly on the trigeminal nerve root entry zone at very high dose (typically 70–90 Gy) to create a controlled radiosurgical lesion that interrupts pain signaling — demonstrating that the same convergent-beam, isocenter-based platform serves both ablative functional targets and conformal tumor/vascular targets, differing only in shot number, size, and prescription dose.
Beyond simply choosing where to place a shot, Gamma Knife planning shapes each shot's dose cloud and protects nearby critical structures by controlling collimator size and by selectively blocking or reweighting individual sectors of the source array — turning a fixed-geometry machine into a highly configurable, patient-specific delivery system.
The original Gamma Knife 4C used four interchangeable hemispheric helmets (4, 8, 14, 16 mm), each pre-drilled with 201 channels; changing collimator size for a shot meant physically swapping the helmet. The Perfexion and Icon models eliminated helmet swapping entirely: the 192 sources are divided into 8 independently controlled sectors, each of which can move to align with a 4 mm, 8 mm, or 16 mm collimator, or retract fully to block that sector's beams — all under robotic and software control, with no manual intervention between shots.
Because each of the 8 sectors approaches the isocenter from a different direction, blocking one or more sectors removes dose contribution from that specific angular range while leaving the rest of the convergent geometry intact. This is used when a shot must sit close to a radiosensitive structure — the optic chiasm and nerves, the brainstem, or the cochlea — that would otherwise receive an unacceptable dose from beams passing near or through it.
By retracting the sectors whose trajectories pass closest to the at-risk structure, the planner sculpts the dose cloud into a non-spherical, protective shape for that individual shot, while sectors approaching from safer angles continue to deliver full dose to the target. Different shots in the same plan can use entirely different sector configurations.
Sector-level control means a single treatment plan can combine a wide-open 16 mm shot in the bulk of a tumor with a partially-blocked 4 mm shot at its margin nearest the brainstem — beam shaping tailored shot-by-shot, not just plan-by-plan.
In addition to collimator size and sector blocking, each shot in a composite plan can be assigned a relative weight, scaling its contribution to the total dose. Central shots deep within a large lesion are often weighted higher; peripheral shots near the margin, or those with partial sector blocking, are typically weighted lower. The final composite isodose surface is the weighted sum of every shot's spherical (or sector-shaped) contribution — allowing the planner to fine-tune not just where dose goes, but how much of it, with a level of control that would be impossible with any single uniform beam.
The final composite plan — the union of every weighted, collimated, sector-shaped shot — is Gamma Knife's therapeutic product: a dose cloud that conforms tightly to an irregular target and then falls to near-background levels within just a few millimeters. This steep gradient, more than raw dose or accuracy alone, is what allows radiosurgery to treat lesions immediately adjacent to structures that cannot tolerate significant radiation.
The gradient index (GI) is defined as the volume of the 50%-of-prescription isodose divided by the volume of the 100%-prescription (target) isodose. A GI near 3 — typical for well-planned Gamma Knife cases — means the dose has fallen to half its prescribed value in a shell only slightly larger than the target itself, i.e., dose collapses from therapeutic to sub-therapeutic within a few millimeters outside the lesion boundary. Lower GI values indicate steeper, more favorable falloff. This is systematically steeper than achievable with linear-accelerator (linac)-based radiosurgery systems using fewer, larger, coplanar or near-coplanar arcs, which is why Gamma Knife retains a niche advantage for very small lesions immediately adjacent to critical structures.
A steep gradient is not a cosmetic detail — it is the physical reason a 4 mm shot can deliver 80+ Gy to the trigeminal nerve root while the adjacent brainstem, only millimeters away, receives a fraction of that dose.
Classic Gamma Knife radiosurgery delivers the entire prescribed dose in one session, exploiting the steep gradient so that the biologically large single-fraction dose is confined almost entirely to the target. Typical margin (prescription) doses range from roughly 12 to 24 Gy depending on lesion type, volume, and location: small brain metastases are often treated near 20–24 Gy (per RTOG 90-05 dose-volume guidelines, with dose reduced as target diameter increases), vestibular schwannomas typically at 12–13 Gy to preserve hearing and facial nerve function, and AVMs typically at 16–25 Gy depending on nidus volume. Larger lesions or those closer to critical structures may instead be treated with hypofractionated regimens (e.g., 3–5 fractions) using the mask-based Icon platform, trading some single-fraction radiobiological potency for reduced risk to adjacent tissue.
For AVMs specifically, Flickinger and Kondziolka developed integrated dose-volume models that combine the probability of nidus obliteration with the probability of radiation-induced complications as functions of margin dose and treatment volume, allowing clinicians to select a prescription dose that maximizes the chance of cure while keeping complication risk acceptably low. These models showed that obliteration probability rises steeply with dose up to roughly 20–25 Gy for small-to-moderate nidi, while complication risk rises with both dose and treated volume — meaning the achievable therapeutic window narrows sharply for larger AVMs, a key reason very large AVMs are often treated in staged, multi-session, or multi-isocenter volume-staged protocols rather than as a single large-volume shot.