When a moving tumor meets sequential pencil-beam spot scanning
Organs move. The lungs, liver, and pancreas shift with every breath, and tumors embedded in or near them move too — typically 5 to 20 mm, occasionally more, along a roughly cyclic path dominated by the superior-inferior (head-to-foot) direction. Any radiotherapy technique that treats a moving target as if it were stationary risks missing part of the tumor or overdosing healthy tissue around it.
Conventional planning uses a CT snapshot acquired over a few seconds, which freezes the tumor at one arbitrary point in the breathing cycle. In reality the tumor sweeps through a volume of positions every few seconds for the entire treatment. Modern planning addresses this by acquiring a 4D-CT — a CT sorted into ~8–10 breathing-phase bins — and defining an Internal Target Volume (ITV) that encompasses the tumor at every phase, or by building motion directly into a "4D robust" optimization.
Even with a correctly sized ITV, the question of how the treatment beam is delivered in time, relative to the target's position in time, remains open. This is the domain of motion management.
Passive-scattering proton therapy and photon IMRT both expose most or all of the treatment field simultaneously on each beam-on segment: a scattered proton field floods the whole aperture in one spill, and an IMRT segment covers its shaped field in one exposure. Pencil-beam scanning (PBS) is fundamentally different — a narrow proton beamlet (a "spot", a few millimeters wide) is steered magnetically to paint the tumor one spot and one energy layer at a time, with thousands of spots delivered sequentially over tens of seconds to minutes per field.
Because delivery is sequential, the tumor keeps moving while the spot map is still being painted. Any given spot lands wherever the tumor actually is at the instant it fires — not necessarily where it was during planning.
A single PBS field can take 10 seconds to 2 minutes to deliver, while a breathing cycle is only 3–5 seconds. That means dozens of independent breathing cycles occur mid-delivery — and each spot "samples" a essentially random phase of that cycle.
Photon dose distributions degrade gracefully under motion — mostly geometric blurring of the field edge, since photon dose deposits gradually along the whole beam path. Protons instead deposit most of their dose in a sharp, narrow Bragg peak whose depth (range) depends sensitively on the density of every tissue the beam has crossed. When breathing shifts lung tissue, bone, or gas pockets into or out of the beam path, the proton range itself shifts — the Bragg peak can land millimeters short or long of the tumor, causing severe local under- or over-dosing that has no photon equivalent. This range effect, layered on top of positional motion, is why interplay is a distinctly bigger dosimetric concern for proton PBS than for photon IMRT or proton passive scattering.
The "interplay effect" is the dosimetric error that arises specifically from the interaction between two independent periodic (or quasi-periodic) processes: the spot-by-spot scanning pattern and the patient's breathing cycle. When these two clocks beat against each other unfavorably, the resulting single-fraction dose distribution can show pronounced hot and cold spots even though the plan itself was correct.
Picture the spot-scanning pattern as one periodic process (sweeping across the tumor cross-section, then stepping to the next energy layer) and the breathing cycle as another. Their interaction produces a "beat" pattern, much like two out-of-tune musical notes: when the scan happens to revisit a location while the tumor is out of phase with where it was planned, that location is under-dosed; when a spot fires while the tumor has drifted into a neighboring healthy-tissue path, dose is deposited outside the target instead.
Because the ratio of scan period to breathing period is rarely a clean whole number, and because breathing itself is not perfectly regular (amplitude and period vary breath to breath), the resulting error pattern is effectively quasi-random from one delivery to the next — which is exactly why it cannot be predicted or corrected after the fact from the plan alone.
Passive-scattering proton fields and photon IMRT segments each expose the (shaped) field essentially all at once per beam-on period, so a phase mismatch mostly blurs or shifts the field edge as a whole rather than creating isolated cold pockets deep inside the target. IMRT also typically delivers dose across many small monitor-unit segments and, critically, across many treatment fractions, which gives motion errors repeated independent chances to average toward the planned distribution over the course of treatment.
PBS proton therapy loses both of these safety nets: the Bragg peak's sharp fall-off means a spot that lands in the wrong place produces a large local dose gradient error (not a soft blur), and — as covered next — the averaging benefit of fractionation is much weaker for protons than for photons.
For photon IMRT, a long-standing assumption has been that random motion errors wash out across a conventional course of 20–35 daily fractions — each fraction samples a different, uncorrelated slice of the breathing cycle, so hot and cold spots in different fractions tend to cancel in the cumulative dose. Studies of proton PBS interplay have shown this assumption does not transfer safely: because of the sharp range-dependent Bragg peak, even averaged interplay errors can leave residual local dose deviations, and increasingly common hypofractionated or stereotactic body proton regimens (as few as 1–5 large fractions) simply do not offer enough independent fractions for averaging to work at all. This is precisely why explicit, per-fraction motion mitigation — not just "trust the fractionation" — is required for proton PBS to moving targets.
Unlike photon IMRT, where interplay is often treated as a secondary, largely self-averaging concern, proton PBS interplay is treated as a primary dosimetric risk requiring active mitigation on every single fraction.
Rescanning is the simplest and most widely deployed interplay mitigation: instead of delivering the full planned spot map once, the same field is delivered multiple times in rapid succession, each time depositing only a fraction of the total dose. Because each pass samples a different point in the breathing cycle, the hit/miss pattern differs pass to pass, and the errors partially cancel when summed.
There are two common implementations. Volumetric rescanning repeats the entire 3D field (every energy layer, every spot) N times before moving on — each full pass takes several breathing cycles, maximizing phase diversity between passes but taking the longest overall. Layered rescanning instead repeats each individual energy layer N times before advancing to the next layer — faster overall since layer switching (energy change) is comparatively slow, but each set of repeats is more localized in depth and can be more sensitive to slower drifts. A related variant, isolayered rescanning, adapts the number of repeats per layer to equalize the delivery time contributed by each layer.
If a single pass leaves a spot with a dose error of roughly ±ε due to interplay, and successive passes sample statistically independent breathing phases, the error on the summed (rescanned) dose falls roughly as ε/√N for N rescans — the same averaging law that governs any repeated independent measurement. In practice the reduction is somewhat less clean because breathing is not perfectly random and the scan pattern itself is deterministic, but empirically 3–10 rescans reduce interplay-related dose heterogeneity from tens of percent down to just a few percent in most clinical lung and liver cases.
Rescanning trades treatment time for dose robustness: doubling the rescan count roughly doubles delivery time for that field, so clinics balance rescan number against total treatment slot length and beam-time throughput.
Rescanning reduces random interplay noise but does not correct systematic range errors from motion-induced density changes, and its benefit shrinks for large motion amplitudes or for hypofractionated regimens where few total fractions are delivered. For this reason rescanning is very often combined with gating or breath-hold (see Stages 4–5) rather than used as a standalone solution for larger-amplitude lung or liver tumors.
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
Respiratory gating takes a different approach from rescanning: rather than averaging out motion errors, it eliminates most of them by only delivering the beam when the tumor is known to be within a narrow, reproducible window of the breathing cycle — typically end-exhale, where lung tumor position is comparatively stable and repeatable from breath to breath.
Because the tumor itself usually cannot be imaged in real time during beam delivery, gating systems infer breathing phase from a surrogate signal. External systems (e.g. optical/infrared marker blocks tracked by camera, such as the widely used RPM system) place a reflective marker on the chest or abdominal wall and track its vertical displacement as a proxy for internal motion. Internal systems use implanted fiducial markers tracked fluoroscopically, or increasingly proton-specific in-room imaging, to verify that the actual tumor position — not just the external surface — matches the surrogate signal. Because chest-wall motion does not always correlate perfectly with internal tumor motion (a phenomenon called baseline drift or hysteresis), correlation between external and internal signals is periodically verified.
End-exhale is favored over end-inhale for two main reasons: the diaphragm and lower lung tend to dwell longer at end-exhale than at peak inhale (a longer, more stable window to gate on), and breathing shows measurable hysteresis — the inhale and exhale paths of the lung/tumor trajectory are not identical, so end-exhale tends to be the more reproducible extreme of the two. A gate window is defined around this phase (e.g., the bottom 30–50% of the amplitude range), and the beam is enabled only while the live surrogate signal is inside that window.
Narrowing the gate window improves positional accuracy (residual intra-gate tumor motion of roughly 1–3 mm is achievable with a tight window) but lowers the duty cycle — the fraction of total time the beam is actually allowed to fire — which lengthens overall treatment time and reduces machine throughput. Typical clinical duty cycles run around 30–50%, roughly doubling to tripling delivery time compared with an ungated field. Widening the gate trades some residual motion blur for a shorter treatment session, so the gate width is chosen per patient based on motion amplitude, regularity of breathing, and clinical time constraints.
Gating does not reduce dose per breath — it simply refuses to deliver dose outside the defined window, so total delivery time stretches to compensate. The dosimetric payoff is that every delivered spot "sees" nearly the same tumor position, restoring much of the spatial accuracy assumed in planning.
In current clinical practice, the most robust moving-target proton treatments rarely rely on a single mitigation strategy. Rescanning, gating, and deep-inspiration breath-hold are frequently layered together, alongside 4D-CT-based robust planning that explicitly accounts for motion rather than treating it as an afterthought.
DIBH asks the patient to inhale to a consistent, coached lung volume and hold their breath for 15–25 seconds while the beam delivers; the process repeats across several breath-holds until the field is complete. During a hold, tumor motion is essentially eliminated (residual drift is typically under 1–2 mm), which removes interplay risk entirely for that segment rather than merely averaging or gating around it. DIBH also has the added radiobiological benefit, in some thoracic and left-breast cases, of increasing lung inflation and moving the heart away from the treatment field. Its limitation is patient factors: DIBH requires reasonable pulmonary function and the ability to reproducibly hold a coached breath, which is not feasible for all patients.
Rather than treating motion purely as a delivery-time problem to be solved with gating or rescanning, modern planning increasingly incorporates it directly into optimization. A 4D-CT captures the tumor and surrounding density across the full breathing cycle; robust optimization then generates a plan explicitly required to maintain adequate target coverage and healthy-tissue sparing across that entire range of motion states (and, for protons, across a range of expected range/setup uncertainties simultaneously). This "motion-inclusive" planning does not replace gating or rescanning at delivery time, but it ensures the plan being delivered is realistic about what the beam will actually encounter fraction to fraction.
Early proton PBS practice was cautious about treating mobile lung and liver tumors at all, given the magnitude of interplay effects seen in early simulation and phantom studies. Since then, a growing body of prospective clinical experience — combining 4D-CT robust planning, rescanning (commonly 5–10×), and gating or DIBH — has demonstrated that PBS proton therapy can be delivered safely and effectively to moving thoracic and abdominal targets, with in-vivo verification (e.g., log-file and imaging-based dose reconstruction) increasingly used to confirm delivered dose matches planned dose on a per-fraction basis. Ongoing multi-institutional trials continue to refine gating thresholds, optimal rescan counts, and criteria for when breath-hold or free-breathing-with-rescanning is the safer choice for a given patient and tumor site.
The clinical consensus is not "avoid moving targets" but "characterize the motion (4D-CT), plan robustly around it, and mitigate interplay actively at delivery" — with combined rescanning + gating/breath-hold now considered the most dosimetrically robust standard for scanned proton therapy of mobile tumors.