PDF
Citation
Print
Introduction
Quality assurance (QA) programs for medical linear accelerators are essential for ensuring the safe and accurate delivery of radiation therapy. The Winston−Lutz (W/L) test is widely regarded as the gold standard for verifying the coincidence between mechanical and radiation isocenters, as recommended by the American Association of Physicists in Medicine (AAPM) Task Group 142 and international guidelines [1,2]. This test is particularly critical for stereotactic radiosurgery (SRS) and stereotactic body radiation therapy (SBRT), where submillimeter geometric accuracy is essential for achieving optimal treatment outcomes [3,4]. The W/L test, first described by Lutz et al. [5], in 1988, uses a small high-density sphere positioned at the machine’s isocenter to evaluate the geometric accuracy of the treatment delivery system. By acquiring portal images at various gantry, collimator, and couch angles, the test quantifies deviations between the radiation field center and the mechanical isocenter. These measurements provide crucial information for maintaining the geometric accuracy required for modern high-precision radiotherapy techniques [6,7]. With the increasing adoption of advanced treatment techniques, such as HyperArcTM (Varian) and single-isocenter multiple-target SRS, the frequency and importance of W/L testing have increased substantially [8]. Current commercial phantom systems present several limitations that affect clinical workflow efficiency and measurement accuracy. Setup complexity is a major concern, as many existing phantoms, such as the BrainLAB W/L phantom (Brainlab AG), require heavy couch adapters and time-consuming alignment procedures, often resulting in setup times exceeding 10 minutes. Furthermore, most commercial phantoms are designed solely for W/L testing and require separate equipment for other routine QA procedures, such as off-axis factor measurements or beam data collection [9]. To address these limitations, we developed a novel multi-purpose phantom device featuring a lock bar fixture system, a precision 3.5 mm tungsten sphere, and a modular design that enables multiple QA applications. The primary objective of this study was to comprehensively evaluate the clinical feasibility and measurement performance of this novel phantom compared with commercially available systems. Secondary objectives included optimizing imaging protocols and field sizes, assessing dose dependency, and validating cross-platform software compatibility.
Materials and Methods
1. Phantom system specifications
The fabricated multi-purpose W/L phantom system (EP phantom, where EP stands for Ewha Phantom) features a precision-machined aluminum lock bar fixture that mounts directly onto linear accelerator couch rails without heavy adapters and incorporates adjustable leveling screws with a quick-release mechanism. It incorporates a 3.5 mm tungsten sphere with a removable support structure and laser-etched alignment marks to ensure optimal radiographic visibility while minimizing artifacts. The system includes an interchangeable mounting system that enables rapid conversion between W/L testing, ion chamber positioning for off-axis factor measurements, buildup adapter mounting for percentage depth dose measurements, and output factor measurements across various field sizes, as shown in Fig. 1a. The SP phantom (where SP stands for Sun Nuclear Phantom; Sun Nuclear) features an 8 mm tungsten carbide sphere embedded in a 6 cm cubic acrylic phantom and provides dedicated W/L functionality, as shown in Fig. 1b. The RP phantom (where RP stands for RaphaRAD Phantom; RaphaRAD Inc.) contains a 10 mm diameter tungsten sphere and offers a multi-purpose platform design with integrated laser alignment verification, as shown in Fig. 1c. Both phantoms were selected as widely used commercial options for comparative evaluation. The RP phantom provides additional capabilities for field size verification beyond its primary W/L testing function.
2. Experimental design and methodology
All measurements were performed on a Varian TrueBeam linear accelerator (Varian Medical Systems) equipped with 6 MV and 10 MV photon beams and an electronic portal imaging device (EPID) with 0.336 mm/pixel resolution at the isocenter and a 150 cm source-to-detector distance. Three distinct imaging protocols were evaluated to determine the optimal balance between measurement accuracy and clinical efficiency (Table 1). For the EP phantom, a systematic comparison of 1×1 cm² and 2×2 cm² field sizes at isocenter was conducted, with each field size evaluated across all imaging protocols through four independent measurement sessions to assess reproducibility. To evaluate the impact of imaging dose on measurement accuracy while adhering to ALARA (as low as reasonably achievable) principles, two dose levels were tested: 2 monitor units (MU) in portal imaging mode for low dose and 20 MU in conventional imaging mode for standard dose. The standard dose of 20 MU was selected based on the imaging protocol recommended by SunMachine (version 3.2, Sun Nuclear) software to ensure compatibility with manufacturer specifications. Image quality metrics, including contrast-to-noise ratio (CNR) and sphere edge detection confidence, were recorded for each dose level. All acquired images were analyzed using two independent software platforms: PipsPro (DoseLab Pro version 7.0, Sun Nuclear), a commercial analysis tool with automated sphere detection algorithms, and SunMachine, a manufacturer-provided platform with integrated QA reporting. Identical DICOM image datasets were processed using both platforms to assess inter-software measurement agreement, algorithm-dependent systematic biases, and detection reliability across varying sphere sizes. The comprehensive protocol comparison aimed to establish standardized imaging parameters that optimize accuracy while maintaining dose efficiency. This multi-platform validation approach enabled a robust assessment of measurement reliability, independent of the analysis software. The systematic evaluation of field sizes and dose levels generated critical data for optimizing clinical W/L testing procedures.
3. Data acquisition and analysis
For each phantom configuration, the following parameters were recorded: 2D deviation representing the maximum radial deviation in the imaging plane (mm), 3D vector magnitude, calculated as √(X² + Y² + Z²) (mm), individual axis positions including X (lateral), Y (longitudinal), and Z (vertical) deviations (mm), gantry sag analysis, showing positional variation as a function of gantry angle, and decomposition of total error into systematic versus random error components. Statistical analysis was performed using SPSS Statistics version 28 (IBM Corporation) and included descriptive statistics, such as mean, standard deviation, and range, for all measurements, as well as independent samples t-tests for phantom comparison with a significance level of P<0.05. Bland–Altman plots were used to assess inter–software agreement, F-tests evaluated the homogeneity of variances between phantom systems, and intraclass correlation coefficients (ICC) were used to evaluate the reproducibility of repeated measurements. The RP phantom was selected as the reference standard because: (1) it is commercially available and widely used in Korea; (2) it offers multi-purpose capabilities similar to those of our EP phantom, unlike the dedicated SP phantom; (3) it represents current standard practice; and (4) using a single reference ensures statistical consistency. Statistical analysis employed two approaches: (1) independent samples t-tests for between-phantom comparisons (Tables 2-5), as different phantoms represent independent measurement populations; (2) paired t-tests were used for within-phantom field size comparisons (Table 6), where the same EP phantom was measured under different conditions, creating a natural pairing. Levene’s test was performed before the independent t-tests, with Welch’s correction applied when equal variances could not be assumed (P<0.05). A comprehensive uncertainty analysis was conducted to quantify all major uncertainty components contributing to W/L measurements. EPID spatial resolution uncertainty was characterized using empirical resolution from repeated analysis of static images (σEPID=0.168 mm), accounting for the subpixel interpolation capabilities of the centroid detection algorithms. Phantom setup reproducibility was determined from ten repeated setups with complete phantom removal and repositioning using laser alignment: EP phantom σsetup=0.12 mm (benefit of lock bar fixture design), SP phantom σsetup=0.18 mm, RP phantom σsetup=0.15 mm. Software algorithm fitting uncertainty was determined from repeated analyses (n=20) of identical image datasets using both PipsPro and SunMachine (σalgorithm=0.05 mm). Mechanical isocenter stability was assessed using 12 months of daily isocenter QA records (σmechanical=0.08 mm). Combined measurement uncertainty was calculated using root-sum-square combination: EP phantom σcombined=0.22 mm (U95=±0.44 mm), SP phantom σcombined=0.26 mm (U95=±0.51 mm), RP phantom σcombined=0.23 mm (U95=±0.47 mm).
Results
1. Overall performance comparison across phantom systems
Analysis of 432 total measurements (144 per phantom system) revealed significant differences in measurement performance across the three phantom systems, with detailed results summarized in Table 2 (summary of phantom performance by protocol and field size). Detailed analysis of individual axis deviations revealed systematic patterns across phantom systems, as presented in Table 3 (individual axis position analysis). The HWL protocol highlights an important distinction between measurement precision and positioning accuracy characterization. While Table 3 shows that the HWL protocol yields the smallest standard deviations on individual axes (indicating excellent measurement reproducibility), Table 2 reports larger mean 3D values for the HWL protocol. This apparent discrepancy is explained by the comprehensive sampling of the HWL protocol. By using 12 gantry angles at 30° intervals with 4 collimator angles, the HWL protocol more thoroughly captures systematic positioning errors and gantry sag effects across the full rotation. The smaller standard deviations reflect high measurement precision (reproducible measurements), whereas the larger mean 3D values indicate more complete characterization of the actual systematic mechanical isocenter deviation. In contrast, the EWL protocol (8 images) and the SWL protocol (12 images at 45° intervals) provide less comprehensive sampling and may miss certain systematic errors, resulting in artificially lower mean 3D values. Thus, the HWL protocol provides the most accurate characterization of the true mechanical isocenter position despite its higher mean 3D values, because it more completely samples the mechanical system’s geometric accuracy across all gantry positions.
2. Field size optimization results
All data presented in Table 6 are from the EP phantom and compare measurement performance between 1×1 cm² and 2×2 cm² field sizes across three imaging protocols (EWL, SWL, HWL). Field size optimization was performed exclusively for the EP phantom as part of its clinical validation. This approach was taken because: (1) the primary aim of our study was to evaluate and optimize the novel EP phantom; (2) commercial phantoms (SP and RP) have fixed, manufacturer-recommended imaging protocols with specified field sizes, and systematic field size variation would deviate from their validated operating protocols; and (3) institutions adopting the EP phantom require evidence-based guidance for optimal imaging protocols. The EP phantom demonstrated consistent improvement in measurement precision with the 2×2 cm² field size compared to 1×1 cm², as detailed in Table 6 (field size impact on measurement precision). The improved performance with the 2×2 cm² field size was attributed to enhanced edge detection capabilities resulting from increased photon statistics and improved signal quality. CNR measurements showed that the 2×2 cm² field exhibited a 28% higher CNR (18.4±2.1) compared to the 1×1 cm² field (14.3±1.8), providing superior sphere visualization. Edge gradient analysis revealed sharper edge profiles with the 2×2 cm² field (gradient: 0.42±0.05 mm−¹) compared to the 1×1 cm² field (0.31±0.04 mm−¹), enabling more precise determination sphere center. Detection confidence scores from the automated algorithms improved by 22% with the larger field size, indicating more reliable and consistent sphere localization. These quantitative improvements in image quality metrics directly translated into reduced measurement uncertainty and enhanced reproducibility for the SP phantom system.
3. Dose-dependency analysis
Evaluation of portal imaging at reduced doses demonstrated maintained measurement accuracy, as detailed in Table 4 (dose-dependency analysis).
4. Software platform agreement
Analysis of identical datasets using both software platforms demonstrated excellent agreement, as presented in Table 5, showing Inter-Software Agreement Analysis. Bland–Altman plots demonstrated no systematic bias between the software platforms, with a mean bias of −0.02 mm (P=0.186) and 95% limits of agreement within the clinically acceptable range of ±0.10 mm. No proportional bias was observed across the measurement range (r=0.082, P=0.442), confirming consistent agreement regardless of measurement magnitude.
5. Protocol optimization
Comprehensive evaluation of the protocols identified an optimal balance for clinical implementation. The EWL protocol achieved an efficiency score of 9.2 out of 10 and an accuracy score of 7.8 out of 10, yielding an overall rating of 8.5 out of 10. The SWL protocol achieved an efficiency score of 7.5 out of 10 and the highest accuracy score of 9.2 out of 10, yielding an overall rating of 8.4 out of 10. The HWL protocol achieved an efficiency score of 6.8 out of 10 and an accuracy score of 8.5 out of 10, with an overall rating of 7.7 out of 10. The efficiency score was based on acquisition time, setup complexity, and workflow integration, whereas the accuracy score was determined by mean 3D deviation, standard deviation, and reproducibility.
6. Reproducibility and long-term stability
Repeated measurements by the same operator demonstrated excellent intra-observer consistency, with the EP phantom achieving the highest ICC of 0.986 (95% confidence interval [CI]: 0.978−0.992), followed by the SP phantom at 0.972 (95% CI: 0.958−0.984) and the RP phantom at 0.968 (95% CI: 0.952−0.982). Measurements performed by different operators demonstrated high inter-observer agreement across all phantom systems. The EP phantom exhibited superior reproducibility with an ICC of 0.978 (95% CI: 0.964−0.988), while the SP and RP phantoms demonstrated ICCs of 0.964 (95% CI: 0.946−0.978) and 0.958 (95% CI: 0.938−0.974), respectively. All ICC values exceeded 0.95, indicating excellent reproducibility for both intra-observer and inter-observer measurements across all three phantom systems.
Discussion
The EP phantom’s enhanced measurement precision, with standard deviations ranging from 0.16 to 0.24 mm, represents a clinically significant improvement for modern stereotactic treatments. Current guidelines recommend maintaining mechanical isocenter accuracy within a 1 mm radius for conventional treatments and within 0.75 mm for SRS/SBRT applications. The EP phantom’s improved precision enables earlier detection of drift, as smaller standard deviations allow identification of systematic changes before they exceed tolerance levels. It also reduces measurement uncertainty by lowering variability in repeated measurements, thereby increasing confidence in QA results. It also enhances baseline establishment through more precise measurements, facilitating better characterization of machine-specific performance. The lock bar fixture design offers substantial workflow improvements, including a 60%–70% reduction in setup time compared to traditional tripod-based systems. It also provides reproducible positioning through direct couch mounting, which eliminates positioning variability. Additionally, it provides multi-purpose functionality, eliminating the need for multiple phantom systems. The 2×2 cm² field size provides optimal photon statistics with approximately four times higher photon fluence than a 1×1 cm² field and a 28% improvement in CNR, enabling more reliable edge detection. Additionally, the penumbra contribution remains less than 15% of the field area, minimizing geometric uncertainty. This field size corresponds to typical SRS cone diameters of 12.5–20 mm, common SBRT field dimensions, and standard small field dosimetry reference conditions, making it clinically relevant. Implementing 2 MU imaging protocols reduces the annual dose to equipment by approximately 80%, minimizing activation and component degradation. It also decreases scattered radiation exposure to personnel during QA procedures and extends EPID lifetime by reducing the cumulative dose. Although 2 MU imaging showed a slightly reduced CNR, the detection success rate remained above 95%, measurement accuracy stayed within 0.1 mm of the standard dose, and image quality was adequate for routine QA when performed by experienced operators. The SP phantom’s performance across different software platforms supports standardized QA protocols across institutions, reliable benchmarking for multi-center trials, and simplified training and competency assessment. This performance meets or exceeds the requirements of AAPM TG-142 mechanical QA specifications. Several limitations should be acknowledged. Testing was focused on 6 MV beams, requiring validation across all clinical energies, and the data were collected from a single institution, necessitating multi-center studies to strengthen conclusions. Additionally, the evaluation was short-term, necessitating assessment of long-term stability over months or years. The sample size was limited, requiring larger datasets for robust statistical power, and testing was performed on a specific linear accelerator model, necessitating validation on systems from different manufacturers. Distinguishing between measurement precision (standard deviation) and positioning accuracy (mean value) is particularly important when interpreting the HWL protocol results. The HWL protocol’s comprehensive angular sampling (30° gantry increments) provides superior measurement precision, indicated by the smallest standard deviations, and more accurate characterization of systematic mechanical positioning errors, reflected in larger mean 3D values. This seemingly paradoxical result actually highlights the protocol’s strength: it reveals the true extent of mechanical imperfections that protocols with coarser angular sampling may underestimate. For clinical QA, this complete characterization is valuable for establishing accurate baseline performance and detecting systematic drift, although the increased measurement time (12 images) must be balanced against workflow efficiency.
Although the EPID’s physical pixel resolution is 0.336 mm at the isocenter, the automated centroid detection algorithms employ subpixel interpolation techniques, including Gaussian fitting and edge-weighted algorithms. Our comprehensive uncertainty analysis indicates that the combined measurement uncertainty for the EP phantom is ±0.22 mm (k=1), supporting the reporting of results to one decimal place (0.1 mm). The observed differences between phantom systems (0.1–0.3 mm range) are comparable to or exceed this measurement uncertainty, supporting the statistical significance of our comparative findings. The superior measurement precision (lower standard deviation) of the EP phantom represents the most robust and clinically meaningful finding. Although absolute mean 3D values are comparable across phantoms (differences within measurement uncertainty), the EP phantom consistently exhibits lower standard deviations (0.1–0.2 mm compared to 0.2–0.3 mm for commercial phantoms). These lower standard deviations reflect true improvements in measurement reproducibility that exceed measurement uncertainty.
Conclusions
This comprehensive evaluation demonstrates that the novel multi-purpose W/L phantom with lock bar fixture provides significant advantages over commercial phantom systems for linear accelerator QA. The EP phantom achieved a 20%–35% improvement in measurement precision, with standard deviations of 0.16–0.24 mm compared to commercial alternatives, and showed statistically significant differences in the SWL protocol (P<0.05). The 2×2 cm² field size with the SWL imaging protocol using 12 beams provided an optimal balance between measurement accuracy and clinical efficiency for routine QA applications. Excellent agreement between analysis software platforms, with mean differences of less than 0.02 mm and ICC greater than 0.97, supports standardization across institutions and vendor platforms. Adequate measurement precision was maintained with 2 MU portal imaging, supporting ALARA principles while maintaining QA quality. The lock bar fixture design reduces setup time by 60%–70% compared to traditional systems, improving workflow efficiency without compromising measurement quality. These results support adopting the SP phantom for routine W/L testing, particularly in facilities performing high-precision stereotactic treatments, where verification of the geometric accuracy is paramount. The phantom’s multi-purpose capabilities, combined with its demonstrated measurement superiority, make it a valuable tool for comprehensive linear accelerator QA programs.
XML Download