Can 150–200 MeV very high-energy electron radiotherapy be validated by phantom dosimetry?

A benchmark gap above 100 MeV

Treatment-planning studies have suggested that very high-energy electron (VHEE) radiotherapy could deliver favourable dose distributions to deep-seated tumours, but the dose models behind those studies had few experimental benchmarks above 100 MeV. This paper addresses that gap directly: it measures 150 and 200 MeV electron beams at CERN's CLEAR facility with radiochromic film and checks whether a Geant4-based GATE Monte Carlo framework reproduces the measurements.

The energy range matters because the paper states that simulations identify roughly 150–200 MeV as necessary to reach the penetration depths needed for deep-seated tumours, while energies below 100 MeV may be insufficient. The validation is therefore aimed at the window in which the clinical case for VHEE is usually argued, not at an arbitrary accelerator setting.

Two beam modes, one fixed scatterer

The study used two beam configurations. A pristine beam was an unscattered Gaussian spot with a measured size (one standard deviation, σ) of about 1.2 mm in air at roughly 52 cm from the exit window. A scattered beam was produced by leaving a 200 μm-thick YAG:Ce screen in the beam path, broadening the spot to about 4.5 mm at 200 MeV and 5.5 mm at 150 MeV at the same distance.

That scattered configuration was a single fixed setting, chosen because it spread dose over larger phantom regions and provided a more divergent test case. In-air spot-size measurements at both energies were fitted and used to derive the source parameters fed into the Monte Carlo model, so the simulation was anchored to measured beam behaviour rather than to nominal machine settings.

Spot size within 1 mm, depth dose within 3%

In air and in a water tank, the simulations reproduced the measured trends in spot size and central-axis percentage depth dose at both energies. At 200 MeV, the paper reports agreement generally within ±1 mm for spot size and 2–3% for percentage depth dose. In water, the measured spot size grew from roughly 4 mm near the entrance to over 10 mm at +45 cm relative to the isocentre, while in air it stayed below 6 mm over the same distance; the model followed both curves.

These are relative-dose and geometric comparisons, not a claim of 1% absolute dose accuracy. The paper separately estimates combined uncertainty for absolute film-versus-Monte-Carlo dose comparison in high-dose, low-gradient regions at 3–5%, dominated by film calibration, scanner variability and inter-film consistency. Readers should keep that distinction: the tight gamma criteria test spatial and relative-dose agreement after global normalisation.

Peak dose tracks delivered charge

At 5 cm depth in water, peak dose rose linearly with delivered charge for both beam modes, across eight independent film measurements per series. The pristine beam reached a 38.9 Gy peak with 2.82 nC, while the broader scattered beam needed 24.59 nC to reach a 22.5 Gy peak. The steeper slope for the pristine beam reflects its more concentrated dose.

This linearity is the practical link between delivery and dose: it supports using measured charge to predict dose and indicates a stable pulsed output. It also shows that upstream scattering can be used to trade spot size against peak dose, which matters for designing delivery schemes rather than for judging biological effect.

High pass rates except one oblique plane

The phantom validation used MARVIN, an anthropomorphic head-and-neck phantom with water-equivalent and low-density foam inserts, irradiated with 200 MeV scattered beams. Agreement was scored with globally normalised 1%/1 mm gamma analysis and a 20% low-dose threshold, meaning points below 20% of the global maximum dose were excluded and the pass rate is the percentage of analysed points with gamma ≤ 1.

Pass rates were 98.7–99.9% for all four single-spot comparisons and both sagittal-plane seven-spot comparisons. The two posterior-oblique seven-spot comparisons were lower: 81.9% for the water-equivalent insert and 97.0% for the foam insert. The authors describe the 81.9% case as a central failure band aligned with the seven-spot delivery direction, appearing in the overlap region of adjacent spots and not reproduced in the single-spot or sagittal comparisons. They interpret it as sensitivity of the narrow multi-spot field to combined experimental and source-model uncertainties rather than as a general model failure.

Dose rates above 10⁸ Gy per second

During the phantom irradiations, the estimated instantaneous per-spot dose rate exceeded 10⁸ Gy s⁻¹, calculated assuming roughly 40 ns delivery per spot (60 bunches at 667 ps spacing). The paper is explicit that this is a beam-delivery metric, not evidence of a biological FLASH effect, and that it differs from a time-averaged plan or field dose rate, which was substantially lower for seven-spot plans because of mechanical phantom translation between spots.

The estimate also carries uncertainties from bunch count, spacing and timing jitter. The authors state these do not affect the dosimetry conclusions but are relevant to translation, and that any biological inference would require dedicated in vivo studies with rigorously controlled dose-rate definitions. The value of this part of the work is that it establishes physical dosimetry under UHDR conditions as a platform for later radiobiology.

Scope ends at the tested energies and geometry

The validation is specific to the experimentally investigated 150 and 200 MeV beams. The authors argue the measurement and simulation methodology should remain applicable above 200 MeV, including a clinically relevant extension to 250 MeV, but say that would require dedicated beam characterisation and benchmarking; GeV-scale beams would need further checks of the Monte Carlo physics models and source description.

The phantom evidence is also concentrated on 200 MeV scattered beams, with single- and seven-spot deliveries at 5 mm spacing. The paper presents the MARVIN comparison as the first detailed validation of a VHEE Monte Carlo framework in an anthropomorphic phantom at these energies, which is why the heterogeneous result carries weight for planning near air-filled cavities. It does not test treatment plans, biological outcomes, or active online dosimetry for pulsed UHDR beams, which the paper names as a remaining translational challenge.

Trust the model most where fields are simple

The decision-relevant pattern is a gradient of confidence rather than a single pass or fail. The validated framework is strongest for single spots and sagittal seven-spot planes, where pass rates stayed at or above 98.7%, and weakest for the posterior-oblique seven-spot field through the water-equivalent insert, at 81.9%. That split, not the headline agreement in air and water, is what a reader should carry forward: the model is a credible anchor for VHEE planning and accelerator design in this energy range, but narrow multi-spot fields crossing heterogeneous material still expose combined experimental and source-model uncertainty that single-spot tests do not reveal.

Sources used in this answer