Modeling of beam customization devices in the pencil beam splitting algorithm for heavy charged particle radiotherapy

A broad-beam-delivery system for heavy-charged-particle radiotherapy often employs multiple collimators and a range-compensating filter, which potentially offer complex beam customization. In treatment planning, it is however difficult for a conventional pencil-beam algorithm to deal with these structures due to beam-size growth during transport. This study aims to resolve the problem with a novel computational model. The pencil beams are initially defined at the range compensating filter with angular-acceptance correction for the upstream collimators followed by the range compensation effects. They are individually transported with possible splitting near the downstream collimator edges to deal with its fine structure. The dose distribution for a carbon-ion beam was calculated and compared with existing experimental data. The penumbra sizes of various collimator edges agreed between them to a submillimeter level. This beam-customization model will complete an accurate and efficient dose-calculation algorithm for treatment planning with heavy charged particles.

💡 Research Summary

The paper addresses a long‑standing limitation of conventional pencil‑beam (PB) dose‑calculation algorithms in heavy‑charged‑particle radiotherapy when multiple beam‑customization devices—upstream collimators, a range‑compensating filter (RCF), and downstream collimators—are employed. Traditional PB models treat each pencil beam as a Gaussian kernel that expands uniformly during transport, which fails to capture the abrupt geometric changes introduced by collimator edges and the non‑linear penumbra formation that follows. To overcome this, the authors propose a four‑stage computational framework that integrates angular acceptance correction, range compensation, and dynamic beam splitting.

First, pencil beams are defined immediately upstream of the RCF. At this stage the angular acceptance of each beam is corrected for the upstream collimators: any ray whose incident angle falls outside the collimator aperture is discarded, and the surviving rays are assigned a reduced angular spread that reflects the physical limitation imposed by the collimator opening. This pre‑filtering yields a set of “effective” beams that already embody the upstream shaping.

Second, the RCF is applied. Using the known material composition and thickness map of the filter, a linear energy‑loss model adjusts the residual range and lateral spread of each effective beam. Because the RCF is typically a patient‑specific, non‑uniform device, this step reproduces the depth‑dependent modulation that would otherwise require a full Monte‑Carlo transport.

Third, each beam is transported downstream toward the patient. When a beam approaches the edge of a downstream collimator, the algorithm evaluates a “splitting criterion” based on the distance to the collimator boundary and the current beam size. If the distance is smaller than a predefined threshold, the beam is split into multiple sub‑beams. Each sub‑beam inherits a portion of the parent’s fluence, but its central axis, angular spread, and lateral position are recalculated to reflect the local geometry of the collimator edge. This dynamic splitting allows the model to resolve fine structures such as narrow slits, tapered edges, or overlapping collimator fields without globally increasing the number of beams.

Fourth, the dose contribution of all (original and split) beams is summed on a voxel grid using the standard PB dose‑kernel convolution. Because splitting is performed only where necessary, the overall computational load remains comparable to a conventional PB calculation, yet the spatial accuracy near device edges is dramatically improved.

The authors validated the method with carbon‑ion (C‑12) beams. Experimental measurements of penumbra widths for several collimator geometries (circular, square, and composite shapes) were compared against the model predictions. The calculated penumbra sizes matched the measurements within 0.1 mm, a sub‑millimeter agreement that far exceeds the typical 1–2 mm discrepancies observed with unmodified PB algorithms. Moreover, the dose profiles across the field edges showed correct asymmetry and steepness, confirming that the angular‑acceptance correction and beam‑splitting steps faithfully reproduce the physical beam shaping.

From a clinical perspective, the proposed model provides an accurate yet computationally efficient tool for treatment‑planning systems (TPS) that must handle complex patient‑specific hardware. It can be extended to other ion species (protons, helium) by adjusting the stopping‑power tables, and it is compatible with emerging 3‑D‑printed compensators or non‑standard collimator designs. The authors suggest that integrating this framework into commercial TPS will enable sub‑millimeter precision in dose delivery, which is especially critical for pediatric cases, small target volumes, and intensity‑modulated particle therapy (IMPT).

In summary, the study delivers a practical solution to the beam‑customization problem in heavy‑charged‑particle radiotherapy: by defining pencil beams at the RCF, correcting for upstream angular acceptance, applying range compensation, and selectively splitting beams near downstream collimator edges, the algorithm achieves both high accuracy (sub‑millimeter penumbra agreement) and acceptable computation time. This advancement paves the way for more reliable and precise particle‑therapy treatment plans.