In modern external beam radiotherapy, the precision delivery of a therapeutic radiation dose to a target volume while minimizing exposure to surrounding healthy tissues is paramount. The total dose received by nontarget tissues, often referred to as the peripheral dose, is an unavoidable repercussion of the treatment process. This peripheral dose originates from three primary sources: radiation scattered within the patient, radiation scattered from beam-shaping components such as collimators and flattening filters, and leakage radiation that penetrates the protective shielding of the linear accelerator (linac) treatment head [1]. The clinical significance of the peripheral dose has grown with the widespread adoption of advanced treatment techniques such as intensity-modulated radiation therapy (IMRT) and volumetric modulated arc therapy (VMAT). These modalities often require substantially higher monitor units (MUs) to deliver a prescribed dose compared to conventional three-dimensional conformal radiotherapy, thereby increasing the relative contribution of head leakage to the overall peripheral dose. For example, Jegal et al. [2] reported an IMRT factor between 2 and 6 for various IMRT and VMAT cases; meanwhile, a prominent shielding reference suggests an IMRT of up to 10 [3]. This elevation in nontarget radiation exposure is a critical concern for long-term patient safety, particularly regarding the risk of secondary radiation-induced malignancies.

Therefore, international standards specify testable leakage limits. IEC 60601-2-1 (“Medical electrical equipment—Part 2-1”) caps photon head leakage in the patient plane (a 2-m-radius circular plane centered at the isocenter, excluding the primary beam) at an average of 0.1% and a maximum of 0.2% of the absorbed dose on the central axis under reference conditions [4]. Outside the patient plane, at 1 m from the electron path, the maximum permitted leakage is 0.5%. Historically, medical physicists verified compliance at acceptance using a “head-wrap” test, as recommended by the International Atomic Energy Agency [5]. However, a 2018 survey conducted by Kisling et al. [6] found that only 53% of respondents reported performing head leakage testing, despite the continuing professional responsibility to detect potential shielding deficiencies.

Previous investigations into linac head leakage have established that its magnitude and spatial distribution are highly dependent on the specific linac model, manufacturer, and beam energy. For instance, a seminal study by Lonski et al. [7] revealed considerable differences in leakage profiles among C-arm linacs from Varian, Siemens, and Elekta, with the Elekta and Siemens units showing, on average, 4–5 times more leakage than a Varian 600C model. However, considering the technological evolution in shielding designs over the last decade, it is necessary to re-evaluate these trends for modern platforms. More recently, with the introduction of novel linac architectures, studies have focused on characterizing these new platforms. Work by Cai et al. [8] and Caravani et al. [9] has detailed the uniquely low and uniform leakage profile of the Varian Halcyon, a ring-gantry system with an integrated design and beam stopper. Despite these valuable contributions, a knowledge gap persists in the literature due to the paucity of studies that provide a direct, side-by-side comparison of the latest generation of linacs from multiple major vendors and across different fundamental architectures (conventional C-arm vs. ring-gantry) using a single, consistent measurement protocol.

This study aims to present a comprehensive, comparative analysis of head leakage radiation, constituting a multi-machine appraisal based on measurements from a diverse suite of seven modern linacs currently in clinical service at Seoul National University Hospital. Utilizing radiochromic film, we quantified and compared the leakage profiles of various Varian and Elekta models to provide a practical dataset that highlights characteristic differences based on manufacturers, linac architecture, and beam energy.

Measurement point IDs follow the numbering shown in Fig. 1 (Points 1–16 for C-arm systems and 1–17 for Halcyon systems). All values are reported as percentages of the 10,000 cGy reference dose. Error bars indicate the standard deviations of the pixel values within the 1×1 cm² ROIs.

At 10 MV, the five C-arm linacs (VB1, VB2, TB1, TB2, and HM) show intermodel and intramodel variability. For example, at the top surface (Point 1), leakage spans 0.166% for Elekta HM to 0.409% for Varian TB1. Fig. 2 highlights these architecture- and location-dependent patterns.

The paired comparisons in Fig. 3 show point-specific sensitivity to energy. Increasing energy from 10 to 15 MV produces a comparable extent of leakage, with minor point-to-point fluctuations, at nearly all points for the three Varian units capable of both energies (VB1, TB1, and TB2), except at Point 1. At Point 1, VB1 increases from 0.257% to 2.044%, TB1 from 0.409% to 1.348%, and TB2 from 0.168% to 1.105%.

In contrast, the ring-gantry Halcyon units (HC1 and HC2) operating at 6 MV FFF exhibit lower and more spatially uniform leakage across all positions (Fig. 4). The largest recorded value is 0.334% (HC1, Point 17) on the in-bore out-of-field surface, with most measurements well below 0.10%, and the two units display closely matching profiles. This uniform suppression is primarily the platform’s low beam energy and its integrated ring-gantry shielding design.

Taken together, these findings demonstrate architecture- and energy-dependent leakage features. The ring-gantry design yields uniformly low values, whereas conventional C-arm heads show more heterogeneous patterns that can be exacerbated at higher energy.

This study provides a comparative dataset on head leakage radiation from a broad range of modern linacs. The analysis of this data reveals distinct patterns related to manufacturer design, beam energy, and fundamental machine architecture, with implications for radiation safety and quality assurance practices.

The data presented in Fig. 2 for the 10 MV photon beam allow for a direct comparison of the shielding effectiveness between the latest generation of C-arm linacs from Varian and Elekta. The Elekta HM demonstrated a comparable leakage profile to the Varian VitalBeam and TrueBeam units. This observation suggests that the trend identified by Lonski et al. [7], where non-Varian machines exhibited higher average leakage than older Varian models, no longer holds for this generation of equipment. The amelioration can be attributed to the difference in the MLC models between MLCi2 (Elekta) and Agility (Elekta). The newer Agility reduced transmission by almost an order of magnitude compared to the previous model [11,12]. The slight differences likely reflect distinct internal component layouts and shielding design philosophies between manufacturers. For instance, the path of the accelerating waveguide, the configuration of the bending magnet, and the materials used in the primary and secondary collimation systems all contribute to the final leakage signature. High leakage was observed at Point 7 for Varian models and at Point 8 for HM. Point 8 of the Elekta machine is located at the superior side collimator level. Due to the geometry of the cover, this is at the corner between the neck along the path of the waveguide and the treatment head. The leakage from the waveguide and the treatment head combined may have caused an elevated reading. On the other hand, Point 7 on the Varian machine is on the patient’s left side of the treatment head at the collimator level. There may be a leakage path through the small gap of the jaw (0.5 cm); however, the asymmetry between Point 6 and Point 7 is not fully understood. While a symmetric design is typically expected, the consistent difference observed across Varian machines suggests a potential asymmetry in internal shielding structure. This specific finding warrants further investigation, possibly through Monte Carlo simulations or consultation with the manufacturer.

In addition to intermodel variability, discrepancies were also observed between units of the same model (e.g., VB1 vs. VB2 and TB1 vs. TB2). The VitalBeam units (VB1 vs. VB2) showed close agreement, whereas a noticeable difference between the TrueBeam units (TB1 vs. TB2) at 10 MV was observed. For instance, differences exceeding 100% (a factor of two) were observed at Point 1 (TB1: 0.409% vs. TB2: 0.168%) and Point 2 (TB1: 0.084% vs. TB2: 0.197%), suggesting potential variations in shielding even within the same model line.

The comparison between 10 and 15 MV beams on the Varian C-arm units (Fig. 3) revealed a localized impact of beam energy. While the physical principle of higher photon penetration at increased energies is well-established, this effect manifested dominantly at Point 1 (Top Surface). This was observed most prominently for the VitalBeam unit VB1, where the surface leakage surged from 0.257% at 10 MV to 2.044% at 15 MV. Conversely, the remaining points (2–16) showed no systematic trend, suggesting the shielding in these areas is robust to this energy increase.

However, it is critical to interpret this value within the proper context of international safety standards. The IEC 60601-2-1 standard specifies leakage limits at defined distances: a maximum of 0.2% in the patient plane (a 2-m-radius circle at the isocenter) and a maximum of 0.5% at a distance of 1 m from the electron path for areas outside the patient plane. The measurements in this study were performed directly on the surface of the linac head covers to map the leakage profile, not at the 1 m distance stipulated for compliance testing. Therefore, the 2.044% value represents a considerable surface “hotspot” rather than a direct case of noncompliance. Radiation intensity decreases with the square of the distance from the source. Assuming the leakage originates from the target that is approximately 20 cm from the outer gantry cover at Point 1 based on the manufacturer’s schematics, extrapolating this surface measurement to the 1 m compliance distance would yield a value within the 0.5% limit. For example, applying the inverse-square law for a source 20 cm from the measurement point suggests the leakage at 1 m would be approximately 2.044% ×(20 cm/100 cm)²=0.08%.

Nevertheless, this finding remains clinically considerable. It identifies a specific shielding vulnerability or design characteristic for this model at 15 MV, located over the bending magnet and primary collimator. The identification of such a notable hotspot, even if compliant at the specified distance, is crucial for the as low as reasonably achievable (ALARA) principle. It contributes to the overall radiation environment in the treatment room, has implications for occupational exposure, and underscores that different machines can have vastly different leakage signatures. This result supports the need for detailed, machine-specific leakage mapping as a critical component of acceptance testing and periodic quality assurance as it provides a more complete understanding of a machine’s radiation profile than standard compliance checks alone.

The data from the Varian Halcyon units (Fig. 4) presents a contrast to that of the C-arm linacs. The Halcyon’s leakage profile is characterized by greater uniformity. Although the maximum measured leakage on either Halcyon unit was 0.334% (at Point 17), most points were below 0.1%. This is consistent with the findings of Cai et al. [8] and Caravani et al. [9], who reported maximum leakage values measured at 1 m from the target of less than 0.04% and 0.03%, respectively. They attribute the superior leakage control of the Halcyon to a combination of its lower beam energy (6 MV FFF) and its fundamentally different architecture, which incorporates comprehensive shielding. A similar pattern to the results of C-arm linacs is observed at Point 1 (top surface), showing a higher leakage compared to other points. On the other hand, Points 14 and 17, located on the in-bore out-of-field surface in superior and inferior directions, are the two highest leakages measured. They were located 3 cm away from Halcyon’s maximum 28×28 cm² field edges. The result may be due to the Halcyon’s dual-layer MLCs design, which features both banks in the x-direction, potentially leading to higher leakage in the y-direction. However, such directional dependence has not been reported in previous studies. In light of IEC compliance, given that the surface is approximately 50 cm from the target based on machine schematics, the estimated leakage at the isocenter plane is 0.083%, which is well within the IEC limit.

Nevertheless, several limitations must be acknowledged in the interpretation of our findings. First, while Gafchromic film offers excellent spatial resolution for mapping leakage profiles, its dosimetric uncertainty can be higher than that of an ion chamber if the calibration and handling protocol is not meticulously followed. Second, this study utilized a discrete 16-point (C-arm) and 17-point (Halcyon) sampling methodology. This provides a leakage profile at key locations but does not represent a continuous surface map. It is possible that the true maximum surface “hotspot” exists in an area between these specific measurement points. Third, this study’s protocol was designed to map the surface leakage profile, not to formally test IEC 60601-2-1 compliance. Therefore, a direct comparison of our surface values to the IEC limits or to previous studies that measured at the 1 m compliance distance is not possible, although extrapolations were provided for context in the manuscript’s discussion. Additionally, all leakage maps were derived from single-irradiation films at each measurement point, without repeated setups. Consequently, we could not explicitly quantify Type A repeatability uncertainties, and the reported values should be interpreted as representative machine-specific leakage signatures rather than population-level statistics. Finally, the data are derived from a single institution’s equipment fleet. While this ensures consistency in measurement protocol, machine-to-machine variations may exist across a broader global population of these linac models.

This study performed a comprehensive and direct comparative analysis of head leakage radiation from a diverse suite of seven modern linacs, employing a consistent film dosimetry protocol. Results demonstrate that leakage profile variations exist, driven by manufacturer design, linac architecture, and beam energy.

While some surface hotspots exceed the numerical value of the IEC 60601-2-1 limit, their extrapolated values at the 1 m compliance distance are likely to be within the standard’s tolerance. Nonetheless, this critical finding highlights the indispensable role of performing detailed, machine-specific, and energy-specific leakage surveys. Such mapping provides a more complete radiological profile of the equipment, revealing shielding features that are important for the ALARA principle and overall radiation safety, beyond standard compliance procedures.