Measurement of microdosimetric spectra of high‐LET particle beams using the ENCORE detector

Abstract

Background

Heavy ion radiotherapy offers distinct advantages over conventional treatments due to its superior dose distribution and enhanced biological effectiveness. However, the microdosimetric characteristics of high linear energy transfer (LET) ions, particularly near the Bragg peak, remain poorly characterized. Most experimental microdosimetry has been conducted at lower LETs, leaving a critical gap in high‐LET microdosimetric data.

Purpose

This study aims to extend microdosimetry to several different high‐LET ions (700–2500 keV/µm) using the novel ENCORE detector.

Methods

Microdosimetry was performed at the Florida State University John D. Fox Accelerator Laboratory using the ENCORE detector, a multi‐sampling ionization chamber filled with low‐pressure tissue‐equivalent gas capable of resolving energy deposition from approximately 200 keV to 3 MeV. ¹²C, ¹⁶O, and ²⁸Si beams were accelerated to 17–55 MeV and delivered to the segmented detector. Energy deposition patterns near the Bragg peak were recorded, and microdosimetric spectra were reconstructed. Monte Carlo was used to model the experimental setup. From each measured and simulated spectra, dose‐mean lineal energy (${\overline{y}}_{\mathrm{D}}$) was calculated and compared.

Results

The 16 strips of the ENCORE corresponded to a tissue‐equivalent thickness of 17.6 µm, enabling resolution of energy deposition immediately proximal and distal to the Bragg peak. In six measurement configurations, the ion beam stopped within this range, allowing precise mapping of the ${\overline{y}}_{\mathrm{D}}$ profile around the peak; in the remaining configurations, the beam fully traversed the detector. Across all beams, difference between measured and simulated ${\overline{y}}_{\mathrm{D}}$ averaged 12%, based on one measurement per strip. Peak measured ${\overline{y}}_{\mathrm{D}}$ reached 920 ± 220, 1240 ± 206, and 2460 ± 620 keV/µm for ¹²C, ¹⁶O, and ²⁸Si beams, respectively.

Conclusion

This study demonstrated that ENCORE can resolve microdosimetric spectra with high spatial and LET resolution across a range of ion energies. Measured ${\overline{y}}_{\mathrm{D}}$ values reasonably agreed with Monte Carlo simulations, validating the potential of ENCORE for position‐resolved microdosimetry and model benchmarking in particle therapy.

1. INTRODUCTION

Heavy ion radiotherapy offers distinct advantages over conventional photon and electron therapy due to its high linear energy transfer (LET) and sharp Bragg peak, which enable precise dose localization and increased relative biological effectiveness (RBE). ¹ , ² , ³ , ⁴ , ⁵ Realizing these clinical benefits requires detailed knowledge of radiation quality, especially in the high‐LET region near the Bragg peak, where biological effects are most pronounced. Microdosimetry provides this insight by quantifying the stochastic energy deposition in microscopic tissue volumes, enabling improved RBE modeling and biologically informed treatment planning. ⁶ , ⁷ , ⁸ , ⁹ The key metric in microdosimetry is lineal energy ($y$), defined as the energy deposited divided by the mean chord length of the sensitive volume. ⁷ Microdosimetric spectra are described by frequency‐weighted ($f(y$)) and dose‐weighted ($d(y)$) probability distributions, which respectively quantify the prevalence and biological relevance of specific energy deposition events. ⁶ , ⁸ From these distributions, parameters such as the dose‐mean lineal energy (${\overline{y}}_{\mathrm{D}}$) are derived, providing deterministic parameters to evaluate radiation quality and support RBE modeling. However, few detectors can resolve energy deposition at the micrometer scale needed to characterize the steep LET gradients that occur through the end of the particle range, and experimental microdosimetric data in this region remain sparse.

Despite the importance of characterizing energy deposition in the high‐LET Bragg peak region, experimental microdosimetric data in this domain remain sparse. Most previous studies have relied on relatively high‐energy ion beams, often hundreds of MeV per nucleon, where the Bragg peak lies well beyond the active volume of the detector. ¹⁰ As a result, microdosimetric measurements are commonly limited to lower‐LET regions such as the entrance plateau or spread‐out Bragg peak, where depth resolution requirements are less stringent and dosimetric conditions more stable. These limitations stem largely from the capabilities of existing microdosimeter technologies. A variety of microdosimeters have been developed to measure stochastic energy deposition at microscopic scales, each offering trade‐offs in terms of geometry, material composition, and spatial resolution. Tissue‐equivalent proportional counters (TEPCs) ¹⁰ , ¹¹ , ¹² , ¹³ and miniaturized TEPCs ¹⁴ , ¹⁵ , ¹⁶ remain among the most established tools for microdosimetry in hadron therapy due to their well‐characterized response and ability to approximate spherical tissue‐equivalent sites within gas‐filled geometries of variable size. However, they offer limited spatial resolution and are often susceptible to pulse pile‐up at clinical beam intensities. ¹⁷ In contrast, solid‐state microdosimeters, including silicon‐on‐insulator and 3D‐silicon detectors, offer higher spatial resolution and are capable of measuring microdosimetric spectra in high‐gradient fields, though they present additional considerations by their non‐tissue‐equivalent composition and fixed sensitive volume. ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ Recent studies have also explored diamond‐based microdosimeters for their radiation hardness and tissue equivalence, ²⁴ , ²⁵ , ²⁶ though these remain in earlier stages of development. While harmonization efforts have been made to compare spectral responses across detector types, ²⁷ , ²⁸ each platform presents distinct advantages and limitations depending on beam type, LET range, and spatial resolution requirements. ²⁴ , ²⁹ , ³⁰ Despite these advances, few detectors can capture fine‐scale energy deposition near the peak of ion range, where LET is maximal but highly variable.

Only a limited number of studies have attempted microdosimetry in the high‐LET Bragg peak region using low‐energy beams. For instance, microdosimetric spectra have been measured in 15–18 MeV proton beams, ¹⁹ , ³¹ but these measurement points remain at relatively low LETs due to the light mass of protons and the fact that measurements were taken in the pass‐through region. Heavier ions such as ⁷Li, ^{1 2}C, ¹⁶O, and ⁴⁸Ti have also been studied at energies between 50 and 170 MeV using silicon‐on‐insulator detectors, ³² but again, the beams traversed the entire detector, preventing localized mapping of the Bragg peak. Thus, high‐resolution microdosimetric measurements in the region of highest LET, immediately proximal and distal to the Bragg peak, are still largely absent from the literature.

This study addressed this gap by performing high‐resolution microdosimetric measurements throughout the Bragg peak using low‐energy ^{1 2}C, ¹⁶O, and ²⁸Si ion beams. We employed the ENCORE detector, a segmented multi‐sampling ionization chamber capable of resolving depth‐dependent spectra at micron increments. Operated at low pressure to simulate tissue‐equivalent site sizes, ENCORE enabled direct measurement of microdosimetric spectra around the Bragg peak in a controlled beam fluence rate of $5.3\times {10}^5$ particles per cm² per second. Results were benchmarked against Monte Carlo simulations using the Tool for Particle Simulations (TOPAS), ³³ , ³⁴ providing a novel experimental dataset that supports improved characterization of high‐LET radiation fields for both radiotherapy and space radiation research.

2. METHODS

2.1. ENCORE detector

Microdosimetric spectra were measured using ENCORE, a multi‐sampling ionization chamber (MUSIC)‐type detector that was designed and constructed at the John D. Fox Accelerator Laboratory at Florida State University. MUSIC detectors have been used to identify highly relativistic particles, ³⁵ , ³⁶ study (α, p) and (α, n) reactions, ³⁷ and measure fusion excitation functions in stable and exotic nuclei. ³⁸ , ³⁹ , ⁴⁰ ENCORE, the first of its kind, is built on this class of detectors but incorporates a segmented anode structure that is well‐suited for microdosimetry. This design enables high‐resolution energy loss measurements and allows the reconstruction of Bragg curves with micrometer‐scale spatial resolution along the beam axis.

Figure 1 shows the schematics of this detector. A 2.54 µm‐thick Havar window separated the gas volume from the high‐vacuum beamline. The active detection medium consists of low‐pressure gas, which serves both as the target and ionization material. Inside the chamber, the anode is segmented into 18 strips, each 1.5 cm long along the beam direction. Of these, 16 central strips (strips 1–16) serve as active channels for sampling energy deposition along the beam path, while the outermost strips (strips 0–17) are used to monitor edge effects and ensure full signal containment. To improve spatial resolution and enable discrimination of complex reaction topologies, each active strip is further subdivided into left and right segments, yielding 32 readout channels.

FIGURE 1.

Schematics of the ENCORE detector. Top panel shows the positioning of the anode, cathode, and Frisch grid relative to the incoming beam. Bottom panel shows the schematics of the individual strips within which energy deposition is measured.

Typical energy loss signals measured with ENCORE range from approximately 200 keV to 3 MeV per strip, with resolution dependent on electronic settings. Signals below this range fall within the noise level of the detector and are not reliably distinguishable. A uniform electric field is applied by a high‐voltage cathode, and a Frisch grid minimizes charge collection distortions to enhance energy resolution. The Frisch grid, positioned between the cathode and anode, serves to isolate the signal induction region from the electron drift region, ensuring that the measured signal reflects only the energy deposited by the ionizing particle. In ENCORE, the grid consists of gold‐coated tungsten wires strung across a printed circuit board frame with 2 mm spacing to ensure uniform field shaping and precise signal normalization. The combination of field uniformity, high‐resolution segmentation, and low‐pressure gas handling made ENCORE uniquely suited for this study.

For this application, ENCORE was filled with a methane‐based, tissue‐equivalent gas mixture (64% CH₄, 32.5% CO₂, 3.1% N₂) at a pressure of 55 Torr, simulating a 1.1 µm thick rectangular tissue‐equivalent site between adjacent strips. This thickness corresponds to the mean chord length of the sensitive volume under the assumption of predominantly forward‐directed radiation, consistent with previous studies. ³¹ The validity of this approximation was confirmed using Monte Carlo simulations, as shown in Figure S4. A gas‐handling system was used to recirculate the gas to maintain constant pressure, temperature, and ionization. Operating at 55 Torr represented the lowest pressure threshold ever achieved in an experiment utilizing the ENCORE detector, significantly extending its operational capabilities. However, because ENCORE is typically operated at pressures of 120 Torr or higher, adjustments were required to optimize detector performance at reduced pressure. The negative cathode voltage was increased to 2–3 kV, approximately twice the usual detector operating voltage used for nuclear physics experiments, to compensate for the lower ionization signals produced at low pressure. Additionally, signal amplification was optimized to ensure accurate resolution of microdosimetric events while mitigating noise.

2.2. Experimental setup

Monoenergetic ion beams of ^{1 2}C (17, 20, and 40 MeV), ¹⁶O (25, 30, 40, and 54 MeV), and ²⁸Si (45 and 50 MeV) were delivered to the ENCORE detector at energies ranging from 17 to 55 MeV at the John D. Fox Accelerator Laboratory. A series of electromagnetic elements in the beamline ensured that a single charge state was delivered for each ion species. The experimental setup in the beamline is pictured in Figure S1, while the nominal stopping power in water and the mean energy of each beam incident on the first active volume strip of the detector (after the Havar window) can be found in Table S1. Given the high LET of these beams near the Bragg peak, beam intensity was carefully controlled to minimize signal saturation and pulse pileup. To achieve this, a tantalum foil was mounted upstream in the beam path as an attenuator. This foil contained micro‐etched holes that allowed only a small fraction of the beam particles to pass through unimpeded, while the majority were stopped by the tantalum, reducing the beam fluence rate to approximately 5.3 × 10⁵ particles per cm² per second. The ENCORE is subject to pile‐up at beam intensities above approximately 10⁴ particles per second, and therefore additional post‐processing steps (described in the Supporting Information) were taken to mitigate pile‐up effects in this study.

The detector was coupled to a high‐resolution digital data acquisition system, as shown in the right panel of Figure S1. Signals induced by ionization events in the low‐pressure gas were collected on the segmented anode, with active strips 1–16 subdivided into left and right segments to improve spatial resolution across the beam axis. Signals from these 32 channels were routed through high‐density LEMO feedthroughs and processed via MPR‐16 multi‐channel preamplifiers, which shaped and amplified the analog charge signals. The signals were further amplified using a gain factor of 64. Signals from the Frisch grid, cathode, and edge strips (0 and 17) were routed via BNC connectors to MPR‐1 single‐channel preamplifiers. The signals from the cathode were amplified with a gain factor of 8. All amplified signals were digitized in real time using three 16‐channel digitizers (Model V1725, CAEN) linked in a daisy‐chain configuration. Based on the electronic configuration and gas pressure conditions used in this study, the resulting strip‐dependent energy resolution of the detector ranged from 0.17 to 0.37 keV per µm per channel.

The data acquisition system was triggered using a fast logic signal derived from the Frisch grid, selected for its sharp timing characteristics. This signal was passed through a fast filter amplifier, constant fraction discriminator, and gate generator to initiate event acquisition. List‐mode data were recorded using the FSUDAQ acquisition system, ⁴¹ which captured timestamped energy values for each individual channel hit. Offline, events were reconstructed using a dedicated event builder that grouped temporally correlated hits across channels. Pulse height signals from the left and right sides of each strip were summed to yield total energy deposition, and artificial discriminator thresholds were applied to suppress background noise and pulse pile‐up, as shown in Figure S3. Energy spectra were then logarithmically re‐binned and calibrated using strip‐specific calibration factors to produce microdosimetric spectra, from which deterministic ${\overline{y}}_{\mathrm{D}}$ values were derived, as described in detail in the Supporting Information.

2.3. Monte Carlo simulations

OpenTOPAS v4.0 ³³ , ³⁴ (based on Geant4 version 11.1.3 ⁴² , ⁴³ ) was used to simulate the experimental setup, including all components of the ENCORE, and to calculate microdosimetric spectra in the 16 strips of the detector. ³³ , ³⁴ Each ion beam was modeled as a Gaussian‐distributed source with an energy spread of 0.085% relative to the nominal energy. To quantify robustness of this value, we varied the 1σ energy spread from 0.01% to 5% for multiple ions and energies (Figure S7). Up to 1%, these variations produced no notable changes in the microdosimetric spectra and <0.5% change in ${\overline{y}}_{\mathrm{D}}$. The practical nominal energy spread of the John J. Fox accelerator beamline measured in this study is approximately 0.025% (1σ), supporting our simulations.

The spatial distribution of the beam followed a Gaussian profile with a position spread of 1.274 mm in both the x‐ and y‐directions, with an elliptical cutoff of 1.5 cm. The angular divergence was also modeled with a Gaussian distribution, using an angular spread of 0.425 mrad in both planes. These beam parameters were chosen as representative values to approximate the expected beam conditions at the entrance of the detector. The upstream tantalum attenuator was not included in the simulation, as it served only as a geometric collimator and did not measurably alter the energy or angular distribution of transmitted particles.

To compute microdosimetric spectra within each of the 16 detector strips, we modified the microdosimetry extension developed by Zhu et al. in TOPAS. ⁴⁴ The original implementation, designed for spherical scoring volumes, was adapted to track energy deposition within the rectangular active volume of the ENCORE detector, aligning with the geometry of its segmented anode structure. This adaptation allowed energy depositions to be scored within a user‐defined cuboid region (rather than the pre‐defined spherical and cylindrical volumes) and converted to lineal energy using a specified mean chord length. For this work, this was assumed to be the strip thickness along the beam axis (1.1 µm), a value that was validated using TOPAS simulations of the particle path length through the strip geometry, as detailed in the Supporting Information. While experimental data were acquired separately for the left and right sides of each strip, simulations scored microdosimetric spectra across the full strip volume without side separation. This approach was to avoid normalization artifacts associated with summing independently normalized spectra, as the microdosimetry extension internally normalizes by the number of entries and bin widths, rather than recording raw event counts.

To improve spectral resolution, the extension was modified to use 1600 logarithmically spaced lineal energy bins from 10^{− 3} to 10⁵ keV/µm, corresponding to 200 bins per decade. For each simulation, ${\overline{y}}_{\mathrm{D}}$ was calculated from the microdosimetric spectrum, and its uncertainty was propagated using statistical moment analysis. Specifically, for each bin, the first and second moments of the bin height ($f(y)$) were calculated across the run histories using a numerically stable algorithm, from which the variance and standard deviation were obtained. These values were then used to compute the uncertainty in the integrated quantities ${\overline{y}}_{\mathrm{F}}$ and ${\overline{y}}_{\mathrm{D}}$, with ${\overline{y}}_{\mathrm{D}}$ uncertainty incorporating contributions from both bin‐wise statistical fluctuations and the propagated uncertainty in ${\overline{y}}_{\mathrm{F}}$. The final standard deviation in ${\overline{y}}_{\mathrm{D}}$ was calculated as the square root of this propagated variance and is reported as the error bar for all simulated values. This method ensured that statistical fluctuations in the spectrum and bin normalization were correctly accounted for in the uncertainty associated with the calculated dose‐mean lineal energy.

To determine the LET of the beam within each strip, an additional scorer was implemented to record the total energy deposited across the entire strip, again treated as a single unit rather than divided into left and right portions. The mean energy deposition was then divided by the simulated tissue equivalent strip thickness of 1.1 µm to obtain the LET for each strip. The variance in LET was calculated based on the distribution of energy deposition across primary histories. Specifically, the standard deviation of the mean energy deposition was computed as the standard deviation reported by TOPAS divided by the square root of the number of contributing histories. This value was then propagated through the strip‐thickness normalization to obtain uncertainty in LET.

Each simulation was run with 10⁷ primary histories, ensuring that the standard deviation in calculated ${\overline{y}}_{\mathrm{D}}$ remained below 0.5%, using 24 processor threads in parallel on the mForge cluster at the National Center for Supercomputing Applications (NCSA, Urbana, Illinois). The physics models used included “g4em‐standard_opt4,” “g4h‐phy_QGSP_BIC_HP,” “g4decay,” “g4ion‐binarycascade,” “g4h‐elastic_HP,” and “g4stopping” to provide a detailed representation of electromagnetic and hadronic interactions. Changes in the effective charge of the ion were accounted for through the dynamic charge model of Geant4, ⁴⁵ which is inherited by the TOPAS framework. A physical electron range cut of 0.05 mm was employed, corresponding to an effective range cut of 4 nm in the low‐pressure gas.

2.4. Results

Microdosimetric spectra measured in the first active strip of the ENCORE detector are shown for each beam configuration in Figure 2. Compared to the Monte Carlo simulations, the measured spectra exhibit a noticeably broader shape, with an increased full width at half maximum, which is primarily attributed to pile‐up signals. While the current analysis applies basic pile‐up rejection, more advanced pile‐up correction techniques are likely needed to fully resolve these effects, as discussed further in the Discussion section. In addition, Figure 3 displays spectra acquired in the sixth active strip for a representative selection of beams, corresponding to locations a few micrometers proximal to the Bragg peak (20 MeV ¹²C and 30 MeV ¹⁶O) and distal to the Bragg peak (50 MeV ²⁸Si). These spectra show similar broadening compared to simulations, though the effect is somewhat less pronounced than in the first strip.

FIGURE 2.

Microdosimetric spectra measured with ENCORE and calculated using Monte Carlo in the first active strip of the detector for each of the different beam configurations assessed in this study.

FIGURE 3.

Microdosimetric spectra measured with ENCORE and calculated using Monte Carlo in the sixth active strip of the detector for a set of three representative beam configurations. For the ^{1 2}C and ¹⁶O beams, this strip corresponds to a position just proximal to the Bragg peak, while for the ²⁸Si beam it is just distal to the Bragg peak. ${\overline{y}}_{\mathrm{D}}$ derived from the spectra are shown in the legend in units of keV/µm.

Figure 4 presents the comparison between the measured and Monte Carlo–simulated ${\overline{y}}_{\mathrm{D}}$ as a function of each strip number and tissue equivalent depth within the detector depth for each ion beam. Measurements were obtained using the ENCORE detector operated at 55 Torr, and simulation results are plotted alongside for each corresponding depth and strip. In six of the beam configurations, measured ${\overline{y}}_{\mathrm{D}}$ increased with depth, reaching a peak near the region of maximum LET before decreasing sharply in the distal portion of the Bragg peak. In the other three (40 MeV ¹²C, 40 MeV ¹⁶O, and 54 MeV ¹⁶O), the beam passed through the detector and ${\overline{y}}_{\mathrm{D}}$ steadily increased throughout all 16 strips. Measured ${\overline{y}}_{\mathrm{D}}$ reached 920, 1240, and 2460 keV/µm across all energies in the respective ¹²C, ¹⁶O, and ²⁸Si beams. The corresponding LET profiles, as a function of tissue‐equivalent depth and strip number, are provided for each beam configuration in Figure S5.

FIGURE 4.

${\overline{y}}_{\mathrm{D}}$ measured and simulated using Monte Carlo within each strip of the ENCORE detector for monoenergetic 12C (17–40 MeV, panel a), 12C (20–54 MeV, panel b), and 28Si (45 and 50 MeV, panel c). Strip number is plotted on the upper x‐axis, while the corresponding tissue‐equivalent depth is plotted on the bottom x‐axis. Error bars represent the 1σ standard deviation in ${\overline{y}}_{\mathrm{D}}$ (<0.5% in Monte Carlo calculated values).

Overall agreement between measured and simulated values was reasonable throughout the detector, with the closest alignment occurring at depths where LET was highest. Some discrepancies were observed near the distal edge of the beam range, where steep LET gradients may have amplified sensitivity to spatial and energy resolution. At the lower lineal energies (below ∼200 keV/µm), signal levels fell below electronic and background noise, making ${\overline{y}}_{\mathrm{D}}$ measurements in this region unreliable with the detector and electronic settings used within this study. Across all beams, the position of the ${\overline{y}}_{\mathrm{D}}$ peak was consistent between measurement and simulation, with the exception of two outlying measurements in the 50 MeV ²⁸Si dataset. The average difference between measured and simulated ${\overline{y}}_{\mathrm{D}}$ across all strips was approximately 12%, with the maximum deviations occurring in high‐LET regions where signal saturation or electronic limitations were most pronounced. When stratified by beam type, the average difference was 11% for pass‐through beams and 14% for beams that stopped within the detector, consistent with increased uncertainty at the distal edge of the Bragg peak.

The largest individual discrepancy was observed for the ²⁸Si beams, shown in panel c of Figure 4. Specifically, in strip numbers three and four for the 50 MeV beam, the measured ${\overline{y}}_{\mathrm{D}}$ values were substantially lower than the corresponding simulation results. These two measurements were clear outliers that significantly increased the spread in calibration factors for their respective strips. While the exact reason these two points deviated remains unclear, possible contributing factors include subtle nonlinearities in electronic configurations or other acquisition‐specific conditions that may have distorted the energy of the primary peak.

To further illustrate the degree of agreement between measurements and simulations, Figure 5 shows the ratio of measured to simulated ${\overline{y}}_{\mathrm{D}}$ values plotted as a function of LET for each data point across all beam configurations. In the top panel, all data points are shown, including those from distal strips where simulated energy deposition is close to zero. In these regions, measured ${\overline{y}}_{\mathrm{D}}$ values were limited by electronic noise and could not be resolved below approximately 150 keV/µm, leading to large ratios that do not reflect meaningful discrepancies. An example of this can be seen in the measured ${\overline{y}}_{\mathrm{D}}$ for strip number nine for the 17 MeV ¹²C beam, where the measured value is several times that of the simulated value. In subsequent distal strips, the measured ${\overline{y}}_{\mathrm{D}}$ shows only minor variation and is entirely attributed to electronic noise.

FIGURE 5.

Relative difference of measured to simulated ${\overline{y}}_{\mathrm{D}}$ as a function of LET for each measured dataset across all ion beams. Grey dashed lines indicate ±0.5 deviation from zero. In the top panel, all data points are shown. In each case where the relative difference exceeded 0.5, the discrepancy occurred at the final strip where energy deposition was recorded. At these distal positions, simulated ${\overline{y}}_{\mathrm{D}}$ values were nearly zero, while measured values were limited by background noise and never dropped below 150 keV/µm. The bottom panel excludes these outlier points to better illustrate the distribution of differences between measured and simulated values. Error bars represent the 1σ standard deviation propagated from measured and Monte Carlo simulated values.

The bottom panel of Figure 5 excludes these distal points (defined as the final strip in which simulated ${\overline{y}}_{\mathrm{D}}$ is greater than zero), allowing for a clearer visualization of the distribution of deviations. Across much of the LET range, the ratio remained within ±0.3, indicating reasonable agreement between measurement and simulation. The remaining variation is likely due to unknown uncertainties in the calibration of the detector, electronic noise, and pulse pileup.

3. DISCUSSION

This experimental setup provided high‐resolution microdosimetric spectra and subsequent deterministic values in a previously unexplored LET range (700–2500 keV/µm), offering critical insights into the energy deposition characteristics of high‐LET ions near the Bragg peak. The use of ENCORE in combination with optimized gas pressures, beam attenuation, and high‐resolution signal processing enabled accurate characterization of heavy ion microdosimetry at a spatial resolution suitable for modeling stochastic energy deposition at the cellular scale. Notably, ENCORE captures the entire Bragg curve in a single run, eliminating the need for repositioning and thereby reducing uncertainties associated with beam fluctuations and detector alignment. These results support the viability of using low‐pressure gas detectors in microdosimetric studies of high‐LET beams, despite the challenging conditions involved.

The measured ${\overline{y}}_D$ values reached 920 ± 220, 1240 ± 206, and 2460 ± 620 keV/µm for ^{1 2}C, ¹⁶O, and ²⁸Si, respectively, which, to the authors' knowledge, represent the first microdosimetric measurements of such high values. The reasonable agreement between measured and simulated microdosimetric quantities (averaging 12%) across a wide energy range provides further confidence in the use of ENCORE for particle therapy applications. Minor discrepancies observed near the distal edge are consistent with low‐signal detection limitations and highlight the importance of ongoing optimization of detector electronics. Although the detector's dynamic range is sufficient to capture lineal energy up to 2700 keV/µm, the precision of such measurements remains limited by current calibration methods and the absence of a more robust pile‐up correction. These findings reinforce ENCORE's potential to benchmark microdosimetric models and refine RBE predictions in high‐LET ion therapy, where the comparison of different RBE models and predictive approaches has become a critical focus. ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹

Accurate RBE modeling in high‐LET regions, where biological response saturates and traditional dose‐based parameters fail to predict effects reliably, relies heavily on mechanistic approaches. Models such as the Microdosimetric Kinetic Model (MKM) ⁵⁰ and Mayo Clinic Florida MKM ⁵¹ , ⁵² , ⁵³ depend on accurate microdosimetric inputs to predict cellular radiation response, especially near the distal Bragg peak where LET sharply rises and dose rapidly declines. The spatially resolved spectra reported in this study enable improved benchmarking and refinement of RBE models under these complex conditions, supporting efforts toward biologically optimized treatment planning. While the present measurements were obtained using a rectangular sensitive volume geometry, a method proposed by Magrin et al. ²⁹ provides a means to convert microdosimetric spectra from slab detectors to those expected for spherical or cylindrical volumes, thereby facilitating comparison with TEPC‐based data and alignment with the assumptions of biological models. Cross‐validation with other detector types, such as tissue‐equivalent proportional counters ¹¹ , ¹⁷ , ⁵⁴ , ⁵⁵ , ⁵⁶ , ⁵⁷ or diamond detectors, ²⁴ , ²⁵ , ²⁶ would further unify measurement approaches across platforms and enhance the integration of microdosimetric data into both radiobiological modeling and space radiation research.

A noteworthy observation in this study was the presence of a small dip in simulated ${\overline{y}}_{\mathrm{D}}$ values just proximal to the Bragg peak. Rather than a monotonic rise and fall in ${\overline{y}}_{\mathrm{D}}$, this slight dip appeared consistently for all beams stopping within the detector (excluding the 40 MeV ^{1 2}C and 40/54 MeV ¹⁶O beams). Monte Carlo investigations revealed that this dip coincided with the depth at which the primary ions’ kinetic energy falls below 5 MeV, causing changes in physics models or transport algorithms within the simulation. At this transition, kinetic energy distributions broaden substantially, contributing to a secondary low‐lineal‐energy peak in the microdosimetric spectrum. Modifying the range cut had no effect, suggesting this phenomenon reflects intrinsic transport physics changes rather than secondary particle artifacts. Importantly, this dip was not observed in macroscale quantities such as LET, which increased smoothly until the Bragg peak (as shown in Figure S5), highlighting the unique sensitivity of microdosimetric quantities to subtle transitions in particle transport and emphasizing the need to understand energy thresholds and model boundaries in Monte Carlo simulations.

While this study demonstrates the capabilities of ENCORE for high‐resolution microdosimetric measurements, several limitations should be acknowledged. The dominant source of measurement uncertainty was the calibration procedure. Because each strip had its own readout electronics, individual calibration was performed by aligning the primary peak of the measured spectrum with Monte Carlo simulations across all beam configurations. The resulting peak ratios were averaged to obtain a final calibration factor per strip, with the standard deviation propagated to measured ${\overline{y}}_{\mathrm{D}}$. This method, while sufficient for proof‐of‐concept, is sensitive to measurement conditions and led to relatively large error bars. Additional uncertainty arose from variations in discriminator thresholds across strips, though a sensitivity analysis showed that ±10% shifts in threshold placement resulted in less than 3% variation in ${\overline{y}}_{\mathrm{D}}$. Future work could significantly reduce overall uncertainty through improved calibration protocols, such as the use of absolute energy references and background subtraction techniques, both of which would improve the robustness of ENCORE's measurements.

There are also complications in the clinical application of ENCORE. As with many microdosimeters, ¹⁰ , ¹⁴ , ¹⁶ it is limited by susceptibility to pileup at high particle flux and by the need for a complex gas handling system and bulky electronics, which together constrain its practicality for routine clinical use. These infrastructure requirements make the detector less suitable for routine QA workflows but rather best suited for exploratory microdosimetry studies or infrequent beamline validation. While the detector performs reliably at beam intensities up to ∼10⁴ particles per second with tissue‐equivalent gas, ³⁸ higher rates induce signal pileup, potentially compromising event‐by‐event analysis. This is particularly important in the context of clinically realistic beam intensities. To address this, pileup discrimination methods such as signal deconvolution ⁵⁸ have been proposed by the community, while delayed coincidence triggering, which enables detection and quantification of overlapping events, has been specifically suggested by the ENCORE developers. ³⁸ Such approaches may help extend the detector's usability to higher dose rates in future applications. These improvements, in conjunction with continued optimization of electronics and threshold settings, will be explored in future studies and are expected to dramatically reduce measurement uncertainty and improve detector performance.

4. CONCLUSION

This study demonstrated that the ENCORE detector is capable of resolving microdosimetric spectra with high spatial and LET resolution across a range of high‐energy ion beams. Representative microdosimetric spectra and derived dose‐mean lineal energy values were compared with Monte Carlo simulations and showed general agreement, particularly in high‐LET regions near the Bragg peak. However, several measured data points exhibited notable uncertainty, primarily due to calibration variability and pulse pile‐up effects. Despite these limitations, the overall relative error averaged 0.12 across all measurement points. These findings support ENCORE's feasibility for position‐resolved microdosimetry in heavy ion fields.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Supporting Information

Hartzell S, Lopez‐Saavedra E, Furutani KM, et al. Measurement of microdosimetric spectra of high‐LET particle beams using the ENCORE detector. Med Phys. 2025;52:e70150. 10.1002/mp.70150