PDF
Citation
Print
Introduction
Accurate tumor localization and targeting are fundamental determinants of therapeutic success in modern radiation therapy [1], and the clinical demand for precision is especially high in particle therapy because of its steep dose gradients and potential to minimize exposure to organs at risk (OAR) [2]. However, respiratory-induced internal organ and tumor motion continue to pose formidable challenges to dosimetric accuracy, particularly in thoracic and abdominal malignancies where tumor excursions can exceed 20 mm [3,4]. Inadequate management of these dynamic anatomical variations can lead to significant discrepancies between planned and delivered radiation dose distributions, potentially undermining treatment efficacy and increasing the risk of toxicity to surrounding normal tissues [5-7].
This high dosimetric sensitivity is exceptionally critical in scanned-beam modalities, such as pencil beam scanning (PBS) proton therapy. A temporal−spatial interference known as the “interplay effect”—arising from the interaction between dynamic beam spot scanning and tumor motion—can introduce substantial distortions to the planned dose distribution [6,8]. This effect can lead to significant target underdosage (cold spots) and unintended overdosage (hot spots) to adjacent normal tissues [9,10].
AAPM Task Group 76 identified tumor displacements exceeding 5 mm as clinically significant [3], and Task Group 324 documented persistent variability in institutional motion management practices [11]. Task Group 290 further highlighted the unique challenges of managing respiratory motion in particle radiotherapy, noting that range uncertainties and Bragg peak shifts further complicate the complexity of motion-inclusive treatment delivery [12]. If respiratory motion is not properly managed, discrepancies may arise between the planned and delivered radiation dose [3,13,14]. Therefore, rigorous quality assurance (QA) processes are essential to ensure that the delivered dose accurately reflects the treatment plan.
The clinical consequences of inadequate respiratory motion QA are substantial. Analyses of phase III radiation therapy clinical trials have revealed that 30%–40% do not require rigorous QA protocols, and such protocol deviations are associated with inferior patient outcomes and decreased trial validity [12]. In particle therapy specifically, the interplay between PBS delivery dynamics and respiratory motion can cause localized dose perturbations exceeding 20%, even when time-averaged dose distributions appear acceptable, underscoring the need for motion-resolved QA capabilities [10].
Existing QA phantoms designed to simulate respiratory motion exhibit substantial limitations that restrict their clinical utility, particularly in particle therapy. Commercial phantoms typically replicate only simplified sinusoidal motion patterns along a single axis and fail to capture the complex, patient-specific three-dimensional trajectories observed in actual respiratory cycles [15,16]. Deformable phantoms incorporating tissue-equivalent materials have been developed to enhance anatomical realism [17-19]; however, they remain constrained by oversimplified geometries, poor reproducibility, and high costs. Critically, these dosimetry-focused phantoms inherently conflate complex physical dose perturbations (e.g., interplay, scattering) with the mechanical and temporal performance of the four-dimensional (4D) delivery system [20]. A significant gap remains in specialized tools capable of isolating and validating the core mechanical and temporal parameters of motion-compensated delivery. These parameters include system latency, internal–external synchronization, and baseline beam-steering stability, which need to be assessed independently of complex dosimetric confounders [21-23].
Internal−external correlation models, which predict internal tumor positions from external surrogate signals, form the foundation of many clinical respiratory tracking systems. However, these correlations exhibit time-dependent variability due to baseline drift and phase shifts, with prediction errors reaching 3.9–5.4 mm under realistic drift conditions of 0.25–0.50 mm/min [24]. Comprehensive QA phantoms that can simultaneously replicate patient-specific internal target motion and coordinated external surface motion are essential for validating these correlation models under controlled, repeatable conditions [25].
This study aims to develop a four-dimensional internal−external respiratory motion phantom (4D IERM Phantom) primarily designed to validate the mechanical and temporal fidelity of 4D scanning beam delivery systems. By simplifying the internal lung geometry to an air cavity, our design isolates the system’s performance from the complexities of heterogeneous dose calculations. The phantom’s objectives are to enable accurate, patient-specific QA of the following parameters: (1) the simultaneous replication of patient-specific internal tumor motion and coordinated external chest surface motion, facilitating comprehensive validation of internal−external correlation models; (2) high-fidelity motion control using patient-derived 4D-computed tomography (CT) and optical surface tracking data, with 0.2 mm spatial resolution and 50 ms temporal resolution; (3) dual dosimetry modes accommodating both multi-active volume ion chambers and the MatriXX 2D planar dosimetry system (IBA Dosimetry) for verifying fundamental beam delivery characteristics (e.g., spot position, profile) and system temporal response (e.g., gating latency) under dynamic conditions; and (4) cost-effective, modular construction using Arduino-based control systems, facilitating customization and lowering barriers to clinical adoption.
Therefore, this study presents the design, development, and validation of the 4D IERM Phantom as a novel standard for mechanical and temporal QA in particle therapy. The phantom’s ability to utilize the MatriXX ion chamber array enables precise verification of planar beam characteristics and temporal delivery accuracy during simulated, patient-specific respiratory motion. By accurately replicating these individualized respiratory patterns, the 4D IERM Phantom is expected to provide a robust method for validating the mechanical and temporal integrity of 4D treatment delivery, thereby enhancing clinical confidence in dose delivery accuracy and minimizing unnecessary radiation exposure to OAR.
Materials and Methods
1. Phantom design
The 4D IERM Phantom consists of two main components: a control unit that manages its operation and a motion unit that simulates the movement of the chest and tumor area Fig. 1. The control unit employs an Arduino Mega 2560 (Arduino) to send signals to the stepper motor drivers, providing the 4D IERM Phantom with both cost-effectiveness and functionality.
2. Control unit
The control unit is the component responsible for managing the phantom’s movements. It consists of a printed circuit board (PCB) based on the Arduino Mega 2560. Through software programming, the Arduino allows for customization of each patient’s respiratory cycle by enabling various velocities and motion patterns. The Arduino’s straightforward interface also simplifies the phantom’s initial setup and future calibration. This system sends signals to the stepper motor via a motor driver, enabling the implementation of desired respiratory patterns in the operation unit.
3. Operation unit
The operation unit is divided into two parts, the chest region and abdomen region, each designed to mimic respiratory movement along specific axes.
1) Chest region
To simulate the rising and falling of the chest during respiration, the chest region employs a single stepper motor to generate 1D motion along the anterior−posterior (AP) direction [26]. Additionally, a pig rib phantom is used to mimic the anatomical structure of the chest wall, enhancing anatomical accuracy.
2) Abdomen region
The abdomen region simulates the complex movement of a tumor target that shifts with respiration, using three stepper motors to mimic 3D motion along the AP, superior−inferior (SI), and left−right (LR) directions. This system enables precise imitation of target motion patterns tailored to each patient, thereby enhancing dose delivery accuracy in particle therapy.
4. Configuration of the 4D IERM Phantom
1) Control unit
The control unit manages the overall control and operation signals for the phantom. It comprises a PCB based on the Arduino Mega 2560 (equipped with the ATmega2560 microcontroller), MD5-HD14 (Autonics, Corp.) and MD5-HD14-3X (Autonics, Corp.) motor drivers, and power supplies (SMPS SPB-240-24 and SPB-120-24, Autonics, Corp.), ensuring stable and efficient operation of the phantom.
2) Chest region
The chest region, responsible for simulating the rising and falling of the chest, includes one A63K-M5913-B (Autonics, Corp.) stepper motor and one E30S4-100-3-T-24 (Autonics, Corp.) rotary encoder. This region is designed to produce 1D motion along the AP direction.
3) Abdominal region
The abdominal region, responsible for simulating abdominal movement, is equipped with two A16K-M569 (Autonics, Corp.) stepper motors and one A63K-M5913-B stepper motor, along with three E30S4-100-3-T-24 rotary encoders for precise control. This setup enables accurate replication of AP, SI, and LR movements in the abdominal region.
5. Specific assignments
1) High-Torque Motor (A63K−M5913−B)
This motor is assigned to the z-axis (Abdomen AP) and the c-axis (Chest AP). The z-axis supports the entire abdominal payload, including the MatriXX or ion chamber, which weighs approximately 20 kg. The c-axis supports the external chest wall mockup (pig ribs) and its mechanism. High torque is essential for maintaining positional accuracy (i.e., preventing step loss) during high-acceleration phases under these significant loads.
2) Lower-Torque Motor (A16K−M569)
Two of these motors are assigned to the x-axis (Abdomen LR) and the y-axis (Abdomen SI). These axes primarily experience frictional loads associated with coronal/transverse motion and are not subjected to the high gravitational load of the detector and upper components, thereby justifying the use of a more economical motor type.
This configuration enables the 4D IERM Phantom to effectively mimic complex patient-specific respiratory motion patterns.
6. Patient data
To accurately validate patient-specific respiratory patterns and tumor movement, data from 72 patients at the National Cancer Center were collected. During the data collection process, a marker with an attached QR code was placed on the patient’s chest to allow natural breathing while recording respiration-related positional data. Markers were positioned on the chest surface, and height changes were recorded using a camera to document chest surface positional data. Planning target volume (PTV) data were extracted from 4D CT scans to obtain positional information for the tumor target [27,28].
The collected respiratory data include positional information at 50 ms intervals, with a minimum spatial resolution of 0.2 mm, recorded in CSV format for analysis. These data are used to adjust the phantom’s motion to match each patient’s respiratory pattern, enabling more realistic simulation.
This study was approved by the Institutional Review Board (IRB) of the National Cancer Center, South Korea (NCC2020-0304). Written informed consent was obtained from all participants.
7. Motion simulation
The 4D IERM Phantom is designed to simulate patient-specific respiratory chest surface distance data and target distance data derived from PTV in 4D CT scans. The phantom’s motion is programmed in 50-ms increments to replicate the actual respiratory cycle. This simulation uses a coordinate system of x, y, z, and c axes (Fig. 2). Specifically, the x-axis represents the LR motion, the y-axis represents the SI motion, and the z-axis represents the AP motion of the abdominal target region. The c-axis represents the AP motion of the external chest surrogate. Maximum displacement and stepwise movement values are summarized in Table 1.
The equations below describe the motor step-to-distance conversions used for each axis, where n denotes the total number of commanded stepper motor steps. The movement per step (Dx(n), Dy(n)) for the x and y axes and the height adjustments in the AP direction (Hz(n), Hc(n)) using the lead screw actuator are given as follows:
Dx(n)=0.006×n(mm) Eq. 1
Dy(n)=0.006×n(mm) Eq. 2
Hz(n)=(0.0184∙n+112.1667)(mm) Eq. 3
Hc(n)=(0.0201∙n+118.6429)(mm) Eq. 4
8. Dosimetry mode
The 4D IERM Phantom offers two dosimetry modes, allowing for the mounting of either an ion chamber with multiple active volumes or a MatriXX device (Fig. 3). These configurations enable precise dose measurements by tracking the movement of the target during respiration, allowing for a direct comparison between planned and delivered dose distributions under patient-specific motion conditions.
9. Validation method
The 4D IERM Phantom was validated using actual patient respiratory data, with assessment focused on displacement, velocity, and acceleration along the x, y, z, and c axes. Pearson correlation analysis was employed to compare the planned and measured positions of the phantom during simulated respiration, thereby assessing the system’s accuracy.
Results
1. Phantom performance
The 4D IERM Phantom simulated patient-specific respiratory patterns by reproducing movements in both the chest and abdomen regions. As reported by AAPM Task Group 76, respiratory motion varies substantially among patients, resulting in diverse tumor motion patterns [3]. To evaluate the phantom’s ability to simulate a range of respiratory patterns, four representative cases were selected from the dataset of 72 patients: (1) a regular sinusoidal pattern, (2) a regular pattern with higher velocity and acceleration, (3) an irregular pattern, and (4) a complex, irregular pattern.
To accurately simulate the actual patient’s respiratory data, a conversion between distance and motor steps was required. For this purpose, the number of stepper motor steps required for movement along each axis was calculated using Eq. 1 to Eq. 4. In Fig. 4, the blue dots represent values converted from actual patient respiratory data, while the red dots indicate the phantom’s actual movements. The difference between the calculated and measured values resulted in a mean absolute error (MAE) of 0.7404 mm, showing that the error remained within 1 mm.
2. Patient data characteristics
Respiratory data were collected from 72 patients at the National Cancer Center. A QR target (60 mm, 65 mm, 75 mm) was placed on each patient’s chest surface to track untrained, irregular breathing patterns. After analyzing these initial respiratory patterns, each patient received a customized guide tailored to their specific breathing patterns. Trained breathing patterns were then recorded with a camera to capture positional data related to each patient’s chest movements. For the abdomen region, target positional data were obtained from 4D CT images, using phase-specific PTV regions to capture the patient-specific target position data. Tables 2 and 3 summarize the collected patient respiratory data.
The phantom’s accuracy in replicating complex, patient-specific motion was comprehensively evaluated using a suite of quantitative metrics, including the Pearson correlation coefficient (r), MAE, root mean square error (RMSE), maximum absolute error (Max AE), and amplitude error. The axis-specific results are presented in Table 4.
Validation was performed using a single, representative patient’s respiratory data set extracted from a 4D CT study. This patient’s complex breathing pattern was used to effectively evaluate the system’s kinematic limits and dynamic stability across all axes simultaneously. The comparison between the target input (calculated distance from the patient’s 4D CT) and the actual output (measured distance from the rotary encoder) for all four axes is presented in Fig. 5. This figure visually demonstrates the high degree of correlation and minimal positional deviation between the commanded and measured positions. The quantitative results of this validation are further detailed in Table 4.
3. Verification of mechanical independence
We conducted a systematic validation to verify the mechanical independence of the abdominal (x, y, z) and chest (c) motion units, confirming that simultaneous operation does not introduce cross-axis mechanical coupling. This was assessed by monitoring the motion fidelity of a reference axis (the y-axis) under increasingly complex dynamic loads. The validation compared the reference axis performance across four distinct operating conditions, ranging from single-axis movement to simultaneous four-axis actuation, with results detailed in Table 5.
The phantom’s closed-loop control system maintains high positional accuracy and stability demonstrated across all operating conditions. Each stepper motor is equipped with a rotary encoder that continuously monitors the actual position of the axis. This continuous feedback mechanism compensates for transient positional loss (step loss), particularly during high-demand kinematic segments of the patient’s respiratory waveform with rapid acceleration. By preventing the accumulation of step loss, the encoder ensures that the system achieves and maintains the high spatial fidelity quantified by the low RMSE values.
4. Engineering validation and load compensation
In the abdominal region, the AP-directional movement carries a load of approximately 20 kg on the z-axis (including the MatriXX), necessitating a support mechanism to handle the weight. To address this, a spring with elastic properties was installed. Fig. 6 demonstrates the effects of using this spring to alleviate the load. Fig. 6a illustrates the abdominal movement without the spring, showing an average movement loss of 50 steps. Figs. 6b, c confirm a reduction in movement loss with the addition of the spring.
Discussion
This study presents the development and validation of a 4D IERM Phantom, designed for QA in dynamic treatment modalities, such as particle therapy. The 4D IERM Phantom has the ability to simultaneously and independently control four axes of motion—three for the abdomen and one for the chest—enabling the replication of complex, irregular, patient-specific respiratory patterns in all directions.
The comprehensive motion fidelity analysis presented in Table 4 strongly supports the phantom’s high reproducibility. The Pearson correlation coefficient (r) exceeded 0.97 across all axes, confirming a high degree of waveform similarity. Critically, the quantitative positional accuracy metrics—RMSE and Max AE—demonstrate the phantom’s precision. RMSE values across all axes remained well below the clinically accepted 5 mm threshold AAPM Task Group 76 [3]. Furthermore, the Max AE for the high-impact z-axis and c-axis (AP motion) remained within approximately 2 mm, demonstrating sub-millimeter level accuracy for most motion patterns.
Regarding mechanical independence—a primary concern for complex 4D motion—the four-axis simultaneous operation validation confirmed the robustness of the system’s design. The change in RMSE between 1-axis and 4-axis operation was minimal (less than 0.8 mm), and the negligible increase in running time confirms that cross-axis mechanical coupling is negligible. This finding verifies that the phantom maintains high dynamic stability even under the most demanding 4D QA scenarios. Additionally, the phantom’s internal control system demonstrated high temporal fidelity, with a command-to-actuation latency of less than 16 ms, critical for accurate gating and rescanning QA.
For transparency, it is important to address a limitation related to the Max AE. The Max AE values in Table 4 represent the maximum instantaneous positional error, which primarily occurs when the command signal imposes high kinematic requirements (specifically, instantaneous velocity and acceleration) that transiently exceed the mechanical limits of the stepper motor and lead screw mechanism. This transient lag results in small, localized positional errors despite the high overall correlation. A clinically relevant mitigation strategy involves providing respiratory coaching or training for patients whose natural breathing exhibits highly non-smooth or abrupt accelerations. Such intervention reduces the kinematic complexity of the input data, helping ensure that the patient’s motion remains within the phantom’s robust operating envelope.
A principal design feature of the 4D IERM Phantom is simplification of the internal medium (lung) to an air cavity, distinguishing it from deformable, tissue-equivalent phantoms [17-19,29,30]. Dosimetric-focused phantoms inherently confound mechanical/temporal delivery errors with complex, motion-induced dose perturbations. Our phantom is expressly designed as a specialized tool for the isolated validation of mechanical and temporal fidelity, independent of these dosimetric confounders.
The limitations of this phantom are defined by its specialized purpose. As it is not a comprehensive dosimetric phantom, it is not intended for validating 3D dose distributions within a heterogeneous, tissue-equivalent medium. The air cavity design, by definition, precludes the quantification of dosimetric distortions resulting from interplay effects within a low-density lung medium. The phantom evaluates the system’s mechanical and temporal response to motion, not the dosimetric consequence of that motion within tissue. Additionally, while the phantom is a robust tool for validating internal−external correlation models, the inherent stability and accuracy of those patient-derived models remain a separate and significant clinical challenge [21,24,25].
Future work will involve comprehensive dosimetric applications that leverage this validated mechanical platform. A quantitative gamma-index analysis using the MatriXX for gated and rescanned PBS deliveries under patient-specific motion represents a key next step. Furthermore, we plan to develop a modular insert for the abdominal unit that allows optional inclusion of solid water, lung-equivalent materials, or other detectors. This addition would create a dual-purpose system, capable of switching between the “mechanical/temporal validation mode” presented here and a conventional “dosimetric validation mode.”
Conclusions
This study detailed the design, development, and validation of a novel 4D IERM Phantom for QA in particle therapy. The system demonstrated high mechanical fidelity in replicating complex, patient-specific respiratory motion for both internal (3D) and external (1D) components.
This study reports the successful development and rigorous validation of the 4D IERM Phantom, which incorporates a four-axis independent actuation mechanism and an Arduino-based control system. Comprehensive motion fidelity metrics, including RMSE and Max AE, confirm that the phantom reproduces patient-specific respiratory patterns with high accuracy (all axes r>0.97) and precision (Max AE contained within approximately 2 mm). The system successfully maintained mechanical independence during simultaneous four-axis operation and achieved an internal time delay of less than 16 ms. These results highlight the phantom’s potential as a next-generation QA tool, capable of independently evaluating motion and dosimetry in the challenging field of motion-managed particle therapy.
XML Download