Characterization of a novel large-area electronic portal imaging device system for a ring-shaped HalosTx linear accelerator in radiotherapy quality assurance

Introduction

Cancer represents a significant global public health challenge (1,2). The American Cancer Society estimated that 2,001,140 new cancer cases would occur in the United States in 2024 (2). Meanwhile, the National Cancer Center of China estimated that there were about 4,824,700 new cancer cases and 2,574,200 new cancer-related deaths in China in 2022 (1). Therefore, as the main treatment method for cancer, the importance of radiotherapy is self-evident (3-5).

In the era of precision radiotherapy, intensity-modulated radiotherapy and volumetric modulated arc therapy (VMAT), which create a high dose gradient between the tumor target and organs at risk (6,7), are extensively used. Therefore, quality assurance (QA) methods that ensure the accuracy of dose delivery are essential. As a method for performing delivered dose QA, in vivo dosimetry (IVD) has significant potential. IVD methods include: (I) point dose measured by tools such as thermoluminescence dosimetry (8), metal oxide semiconductor field-effect transistor (9); (II) electronic portal imaging device (EPID)-based two-dimensional (2D) or three-dimensional (3D) dose reconstruction (9-11); (III) transmission dosimeter-based beam measurement (12,13); and (IV) log file-based dose reconstruction (14). Among them, EPID-based IVD has received considerable attention in recent years (9,15-18).

EPID-based dose QA can be divided into non-transit dosimetry for pretreatment verification methods and transit dosimetry for IVD. The transmission beam fluence maps measured by EPID, which contain extensive information, can be used for 2D dose analysis (19-21) or 3D dose reconstruction (22-24). Currently, several commercial software programs for EPID-based IVD have been developed, such as Dosimetry Check (25) (MathResolutions LLC, Maryland, USA), SOFTDISO (26) (Best Medical Italy, Chianciano, Italy), and PerFRACTION (27) (Sun Nuclear Corporation, Melbourne, FL, USA). These tools are usually integrated with the fluence maps measured by currently available EPID systems described in the American Association of Physicists in Medicine Task Group 307 report (15). Due to their limited detector area, conventional EPID systems (15) struggle to fully cover the whole fluence maps of patients with extensive planning target volumes (PTVs) during IVD, particularly in cases such as breast or esophageal cancer. However, the HalosTx EPID addresses this limitation with its active area of 65 cm × 61 cm, enabling full-field monitoring for all clinical plans.

Recently, the computed tomography (CT)-linear accelerator (linac) HalosTx (United Imaging Healthcare, Shanghai, China) introduced a novel large-sized EPID system. This system contains several advanced features, including a detector effective area of 65 cm × 61 cm, a probe matrix of 4,608×4,352, a pixel spatial resolution of 0.14 mm, and an image acquisition speed of 40 frames per second. Equipped with this large-sized EPID system, comprehensive QA can be performed, including phantom-less pretreatment patient-specific quality assurance (PSQA), real-time 2D IVD during treatment, and post-treatment Monte Carlo-based 3D dose reconstruction and verification. This provides robust QA solutions for most patients, particularly those with extensive PTVs.

Undoubtedly, the above-mentioned features could pave the way for novel EPID-based applications in radiotherapy, such as the development of real-time 3D IVD verification or real-time dose guided radiotherapy. However, this novel EPID system needs to be thoroughly evaluated and characterized before its potential applications can be explored. Thus, this study conducted and assessed characteristic tests for the EPID system and simulated common treatment errors using solid water and anthropomorphic phantoms to evaluate the large-area advantage and the sensitivity of the system.

Methods

EPID system on CT-linac HalosTx

All the tests and measurements were performed on the newly released ring-shaped linac HalosTx, developed by United Imaging Healthcare. This device features a ring-shaped linac, a diagnostic-quality 80-slice helical CT imager, and a 120-leaf multileaf collimator (MLC), utilizing 6 megavolt (MV) photon beams.

Figure 1 illustrates the content and workflow of the study. Basic characteristics, MV imaging quality, field coverage, and sensitivity verification tests were performed for the large-sized EPID with a detector effective area of 65 cm × 61 cm. To investigate the sensitivity of the HaloxTx EPID in detecting various treatment errors such as simulating machine-related errors (e.g., jaw position errors, MLC leaf position errors, output errors, and collimator rotation errors) and patient setup errors, baseline images were first acquired for each case without introducing any errors. These baseline images were used as reference data. Subsequent images captured after introducing specific errors, were compared against the baseline images using distance-to-agreement (DTA), dose difference (DD), and gamma analysis to assess the sensitivity of the EPID system.

Figure 1 Study workflow. Basic characteristic tests, MV imaging quality tests, field coverage tests, and sensitivity tests of the EPID were performed. EPID, electronic portal imaging device; FPS, frames per second; MLC, multileaf collimator; MV, megavolt.

Basic characteristic tests of the EPID

Basic characteristic tests of the EPID were conducted, including short-term repeatability, dose-response linearity, the influence of gantry angle on dose, and dose-rate dependencies. The tested items and their corresponding test conditions are shown in Table 1; all the tests in this section were performed with the treatment couch positioned outside the gantry.

MV imaging quality tests of the EPID

The acquisition of MV portal imaging for applications such as image registration and machine QA represents the primary use of the EPID. Accordingly, the MV imaging quality of the EPID was evaluated, including scaling, spatial resolution, uniformity, contrast, and noise. As shown in Figure 2A, the MV-QA phantom (Sun Nuclear Corporation) was placed on the treatment couch and irradiated using a field size of 25.2 cm × 25.2 cm, a gantry angle of 270°, and a monitor unit setting of 0.2 monitor unit (MU). The resulting images were imported into SNC Machine (version 1.8.1, Sun Nuclear Corporation) for automated analysis. Specific modules in the phantom image, as shown in Figure 2B-2F, were extracted and evaluated to derive the corresponding results.

Figure 2 MV imaging quality tests. (A) The MV-QA phantom. (B-F) The orange regions of interest indicate the modules used for image analysis, corresponding to scaling, spatial resolution, uniformity, contrast, and noise test, respectively. lp, line pair; MV, megavolt; QA, quality assurance.

Field coverage test of the EPID

The field coverage of the HalosTx EPID system for various treatment plans was evaluated by calculating the ratio of plan beam coverage that could be fully monitored in vivo using the EPID. For comparison, the EPID systems of Vital Beam and Versa HD were also tested. Among them, the EPID of HalosTx has dimensions of 65 cm × 61 cm and a source image distance (SID) of 153.4 cm; the EPID of Vital Beam has dimensions of 40 cm × 40 cm and a SID of 150 cm for IVD verification; and the EPID of Versa HD has dimensions of 41 cm × 41 cm and a SID of 160 cm. Planning beam fields were randomly selected for patients with head and neck (509 fields), breast (249 fields), chest (545 fields), abdomen (200 fields), and pelvic (184 fields) tumors. The mean, median, and maximum values of the beam field sizes in the Y direction [superior-inferior (SI)] were recorded. Based on the EPID sizes of the three linacs, calculations were performed to determine whether the EPID systems could fully cover the target areas.

Sensitivity tests of the EPID

Sensitivity tests were conducted to evaluate the ability of the EPID to detect jaw position errors, MLC leaf position errors, output errors, collimator rotation errors, and patient setup errors. The results were analyzed by DTA, DD, and gamma analysis. The introduced machine-related errors are summarized in the first four rows of Table 2.

For the jaw position errors, a 10 cm ×10 cm field with 100 MU and 0° gantry was delivered. Position errors of 1, 2, 3, and 4 mm were introduced to the X1, X2, Y1, and Y2 jaws, respectively.

For the MLC leaf position errors, a 5 cm × 13 cm field with 100 MU and 0° gantry was delivered. Position errors of 1, 2, 3, 4, 5, and 6 mm were introduced; for the left bank, starting from the third leaf, the errors increased by 1 mm every three leaves; for the right bank, starting from the fifth leaf, the errors increased by 1 mm in groups of four leaves.

For the output errors, 10 cm × 10 cm fields with 0° gantry and MU values of 98.5, 99, 99.5, 100, 100.5, 101, and 101.5 were delivered.

For the collimator rotation errors, a 10 cm× 10 cm field with 100 MU and 0° gantry was delivered at collimator angles of 0°, 1°, 2°, and 3°.

For the patient set-up error simulation, CT images of a female thorax phantom were acquired using the GE Revolution ES 128-row CT scanner (General Electric Medical Systems, Waukesha, WI, USA). A PTV of breast cancer was simulated on the image set, and a VMAT plan with a prescription dose of 50 Gy in 25 fractions was optimized. The plan included two partial arcs rotating from 295° to 140° clockwise [treatment beam 1 (TB1]] and from 140° to 295° counterclockwise [treatment beam 2 (TB2)]. The phantom was set up on the treatment couch, the plan was delivered, and the baseline EPID fluence map without set-up errors was acquired after set-up error correction through fan-beam CT scanning. Setup errors were introduced by moving or rotating the five-dimensional treatment couch, as detailed in the last four rows of Table 2. Error-induced fluence maps were compared to the baseline using DTA/DD.

Results

Basic characteristics of the EPID system of HalosTx

The basic physical characteristics of the EPID system, including short-term repeatability, dose-response linearity, the influence of gantry angle on dose, and dose-rate dependencies were evaluated. In terms of the short-term repeatability, using the first measurement as a reference, the deviation for subsequent measurements remained below 0.3%, demonstrating excellent repeatability (Figure 3A). The EPID exhibited high linear responsiveness between MU and corresponding dose, with deviations remaining below 0.6% across MU values ranging from 5 to 2,000 (Figure 3B). In relation to the influence of gantry angle on dose, using the measurement at 0° gantry angle as a reference, the angular dependence remained below 0.8% for all angles. The maximum deviation of 0.757% occurred at 180° (Figure 3C). Under different dose rates, the deviations in the measured dose ranged from 0.015% to 0.372%, with an average deviation of 0.220%±0.139%. The measured dose peaked at 400 MU/min and gradually decreased as the dose rate increased (Figure 3D).

Figure 3 Basic physical characteristics of the EPID. (A) Short-term repeatability. (B) Dose-response linearity. (C) The influence of gantry angle on dose. (D) Dose-rate dependence. EPID, electronic portal imaging device; MU, monitor unit.

MV imaging quality tests of the EPID

The various tests for MV imaging quality were performed by different modules in the MV-QA phantom. The results of MV imaging quality tests are shown in Figure 2B-2F; the orange regions of interest represent the detection areas corresponding to each test. The results of scaling, spatial resolution, uniformity, contrast, and noise were 0 mm, 0.31 lp/mm, 99.39%, 0.74, and 161.33, respectively.

Field coverage of the EPID system for various treatment plans

The field coverage of the EPID systems of HalosTx, Vital Beam, and Versa HD for different treatment plans was evaluated, and the results are summarized in Table 3. The HalosTx EPID successfully monitored all beam fields across various treatment plans, confirming its broad clinical applicability. Conversely, the EPID systems of Vital Beam and Versa HD were applicable to over 90% of the beam fields for most plans but could only monitor 60% of the beam fields for breast cancer plans.

Machine-related sensitivity of the EPID system of HalosTx

The sensitivity of the HalosTx EPID system to machine-related errors is summarized in Table 4. For the jaw position error analysis, the DTA setting value was decreased from the introduced values until the error could be detected. The introduced error of 4 mm was detectable when the DTA value was set to 3.8 mm, indicating a sensitivity of 0.2 mm (Figure 4A). For the MLC leaf position errors analysis, the DTA was used to analyze the results. For an introduced error of 1 mm, the error was detectable at a DTA value of 0.8 mm, indicating a sensitivity of 0.2 mm (Figure 4B). For the collimator rotation error analysis, the DTA was also applied. For errors of 1°, 2°, and 3°, the detectable DTA values were 1.2, 2.1, and 3.3 mm, corresponding to equivalent angles of 0.9°, 1.57°, and 2.53°, respectively. The sensitivity was determined to be less than 0.5° (Figure 4C). For the machine output errors, the DD was used to analyze the results. For introduced errors of 0.5%, 1.0%, and 1.5%, the DD values were 0.6%, 1.1%, and 1.6%. Therefore, the sensitivity was determined to be 0.1%.

Figure 4 Machine-related sensitivity tests of the EPID. (A) Jaw position test. From left to right: EPID-measured baseline image, EPID image with induced jaw position errors (X1 =1 mm, X2 =2 mm, Y1 =3 mm, Y2 =4 mm), and the difference image obtained by subtracting the baseline image from the delivered image. (B) MLC leaf position test. From left to right: MLC position for baseline image and error image, EPID-measured baseline image, EPID image with an induced MLC leaf position error of 1 mm, and the difference image obtained by subtracting the baseline image from the delivered image. (C) Collimator rotation test. From left to right: EPID-measured baseline image, EPID image with an induced collimator rotation error of 3°, and the difference image obtained by subtracting the baseline image from the delivered image. EPID, electronic portal imaging device; MLC, multileaf collimator.

Sensitivity of breast phantom setup errors for the EPID system of HalosTx

A female thorax phantom was used to simulate a left-sided breast cancer treatment plan (Figure 5A,5B). Translational errors of 1 mm in the left-right (LR), anterior-posterior (AP), and SI directions, rotational errors of 0.5° and 1°, and various error combinations were introduced to evaluate the sensitivity of the EPID to setup errors. To enhance the analysis, the gamma passing rates for TB1 and TB2 of the breast plan were calculated separately. Figure 5C illustrates the average gamma passing rates for both individual analyses of TB1 and TB2, as well as their combined analysis. This conventional approach did not distinctly highlight the impact of the introduced setup errors. To further improve the sensitivity of the EPID to setup errors, this study proposed angle-dependent gamma analysis, where each beam was further divided and analyzed at 60° gantry angle intervals (Figure 5B).

Figure 5 Sensitivity tests of setup errors for the EPID. (A) The female thorax phantom. (B) Illustration of the beam fields of the breast cancer treatment plan and angle-dependent gamma analysis. (C) Gamma passing rate of the two beam fields of the breast cancer treatment plan under various introduced setup errors. Gamma criteria: 1%/1 mm. (D) Gamma passing rate of TB1. Gamma criteria: 1%/1 mm. (E) Gamma passing rate of TB2. Gamma criteria: 1%/1 mm. AP, anterior-posterior; EPID, electronic portal imaging device; GA, gantry angle; GPR, gamma passing rate; LR, left-right; SI, superior-inferior; TB1, treatment beam 1; TB2, treatment beam 2.

Figure 5D,5E illustrate the gamma passing rates for different measurement ranges under a 1%/1 mm gamma tolerance criterion. The gamma passing rates exhibited significant variation across different measurement ranges. For angle ranges of 295°–356° and 356.1°–56° (TB1) and 78.1°–18° (TB2), the passing rates showed minimal variation across setup errors. Conversely, significant differences in passing rates were observed for angle ranges of 56.1°–140° (TB1) and 140°–78° (TB2), highlighting the sensitivity of the EPID to setup errors in these angle ranges.

The phantom test demonstrated that the HalosTx EPID system exhibited high sensitivity to setup errors, particularly when the unique angle-dependent gamma analysis method was employed.

Discussion

EPID-based QA has become the most widely used QA tool in radiotherapy due to its convenience, accuracy, and, most importantly, its versatility, which includes non-transit dosimetry for pre-treatment verification and transit dosimetry for IVD. With the rapid advancement of radiotherapy technologies, precision radiotherapy has progressed significantly, inevitably raising the performance requirements for EPID systems.

However, current EPID systems often suffer from limited detector effective areas, frequently failing to fully perform pre-treatment PSQA and IVD for clinical plans involving long PTVs, such as those for breast or esophageal cancer with extensive SI dimensions. For instance, in our department, the EPID systems of Vital Beam and Versa HD, two widely used linacs, could only monitor 60% of the beam fields for randomly selected breast cancer plans with the SID of 150 and 160 cm, respectively. This limitation implies that the dose accuracy for the remaining 40% of the beam fields cannot be guaranteed, posing potential risks to patients. Conversely, the large-sized EPID of HalosTx successfully monitored all beam fields across the selected plans, regardless of the treatment site, demonstrating its ability to meet the stringent requirements of pre-treatment QA for a wide range of clinical plans.

As a widely recognized IVD tool in clinical practice, EPID systems must exhibit high sensitivity to various types of errors. However, before conducting sensitivity testing, it is essential to verify the stability of the novel large-panel EPID to ensure its reliability in detecting deviations (15). Therefore, tests were performed to assess the short-term stability, dose linearity, the impact of gantry angle on dose, and dose-rate dependence. The results revealed that fluctuations in all the tested parameters remained within 0.8%, confirming the high stability and robustness of the system. Meanwhile, the results of the MV imaging quality test demonstrated that the EPID possesses favorable image quality. Nevertheless, long-term stability requires further investigation to ensure reliability in clinical workflows.

Machine-related errors, including jaw position, MLC position, linac output, and collimator rotation errors, were then introduced. The EPID demonstrated sensitivities of 0.2 mm for jaw position, 0.2 mm for MLC position, 0.1% for accelerator output, and 0.5° for collimator rotation. These results align closely with findings reported by Zhuang and Olch (28), while exhibiting slightly higher sensitivity to MLC position and linac output errors. This outcome confirms the superior sensitivity of the EPID and its potential for identifying machine-related deviations through EPID fluence analysis, such as radiomics-based EPID analysis for delivery error classification (20), supporting its role in robust QA protocols.

Although EPID demonstrates high sensitivity to machine-related errors, its application in 3D transit dosimetry shows limited sensitivity to patient setup errors (29). The real-time detection of these errors during treatment is more challenging. Based on a rapid Monte Carlo algorithm, the HalosTx EPID system features a real-time monitoring function, termed the EPID in vivo, which calculates the image sequence acquired by the EPID and compares it with images predicted using the origin treatment plan. This capability enables the real-time monitoring of deviations during treatment. Translational errors and rotational errors were simulated in this study. To avoid potential biases from dose calculation algorithms, EPID images acquired under error-free setup conditions served as the baseline for comparison, rather than relying on computed dose distributions. To improve the sensitivity to setup errors, this study proposed angle-dependent gamma analysis. This represents a significant methodological advancement in EPID-based QA, addressing a critical limitation of conventional gamma evaluation frameworks that treat the entire beam field as a homogeneous entity.

Traditional dose analysis methods (Figure 5C), while widely adopted for planar or volumetric dose comparison, often average spatial and angular dependencies of errors, potentially masking localized deviations that occur in specific gantry angle ranges. Conversely, the angle-dependent approach partitions the EPID-acquired fluence maps into angular segments (e.g., 60° intervals in this study) and performs gamma analysis in each segment. This method leverages the geometric relationship between the beam’s eye view and patient anatomy to enhance sensitivity to setup and machine errors that manifest directionally. For instance, in the simulated breast cancer plan, translational errors in the AP and LR directions exerted pronounced effects on gamma passing rates within the 56.1°–140° (TB1) and 140°–78° (TB2) angular ranges, corresponding to tangential beam geometries. This angular specificity arises because setup errors perpendicular to the beam axis (e.g., couch shifts in AP/LR directions) disproportionately alter the beam path length and fluence distribution in specific angles, which are magnified when analyzed regionally rather than globally. Such directional sensitivity is unattainable with conventional dose analysis, which dilutes these effects across the entire field. Therefore, the necessity of the proposed angle-dependent gamma analysis for IVD is self-evident.

This study had several limitations. First, only short-term performance tests of the EPID were conducted, and its long-term stability remains to be further explored. Second, the sensitivity to setup errors was evaluated based on phantom studies. For practical clinical error detection, clinical treatment data must be collected and site-specific, EPID-based IVD protocols for different treatment sites must be developed for further analysis.

Conclusions

This study comprehensively evaluated the novel EPID system of HalosTx. The system demonstrated superior performance in stability (deviations <0.8%), imaging quality, and error sensitivity (0.2 mm/0.1%/0.5°), and had broad field coverage (100%), enabling real-time QA for complex radiotherapy plans.

Acknowledgments

None.

Footnote

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1908/dss

Funding: This work was supported by National Natural Science Foundation of China (grant Nos. 62576228 and 12475348) and Sichuan Science and Technology Program (grant No. 2024NSFSC1789).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1908/coif). D.G. is from United Imaging Healthcare. The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Han B, Zheng R, Zeng H, Wang S, Sun K, Chen R, Li L, Wei W, He J. Cancer incidence and mortality in China, 2022. J Natl Cancer Cent 2024;4:47-53. [Crossref] [PubMed]
2. Siegel RL, Giaquinto AN, Jemal A. Cancer statistics, 2024. CA Cancer J Clin 2024;74:12-49. [Crossref] [PubMed]
3. Chen CP. Role of External Beam Radiotherapy in Hepatocellular Carcinoma. Clin Liver Dis 2020;24:701-17. [Crossref] [PubMed]
4. Merie R, Gee H, Hau E, Vinod S. An Overview of the Role of Radiotherapy in the Treatment of Small Cell Lung Cancer - A Mainstay of Treatment or a Modality in Decline? Clin Oncol (R Coll Radiol) 2022;34:741-52. [Crossref] [PubMed]
5. Zhang S, Zeng N, Yang J, He J, Zhu F, Liao W, Xiong M, Li Y. Advancements of radiotherapy for recurrent head and neck cancer in modern era. Radiat Oncol 2023;18:166. [Crossref] [PubMed]
6. Otto K. Volumetric modulated arc therapy: IMRT in a single gantry arc. Med Phys 2008;35:310-7. [Crossref] [PubMed]
7. Park JM, Kim K, Chie EK, Choi CH, Ye SJ, Ha SW. RapidArc vs intensity-modulated radiation therapy for hepatocellular carcinoma: a comparative planning study. Br J Radiol 2012;85:e323-9. [Crossref] [PubMed]
8. Kry SF, Alvarez P, Cygler JE, DeWerd LA, Howell RM, Meeks S, O'Daniel J, Reft C, Sawakuchi G, Yukihara EG, Mihailidis D. AAPM TG 191: Clinical use of luminescent dosimeters: TLDs and OSLDs. Med Phys 2020;47:e19-51. [Crossref] [PubMed]
9. Zeng Y, Li H, Chang Y, Han Y, Liu H, Pang B, Han J, Hu B, Cheng J, Zhang S, Yang K, Quan H, Yang Z. In vivo EPID-based daily treatment error identification for volumetric-modulated arc therapy in head and neck cancers with a hierarchical convolutional neural network: a feasibility study. Phys Eng Sci Med 2024;47:907-17. [Crossref] [PubMed]
10. Kirby J, Chester K. Automation to facilitate optimisation of breast radiotherapy treatments using EPID-basedin vivodosimetry. Phys Med Biol 2024;
11. Esposito M, Baldoni R, Bossuyt E, Bresciani S, Clark CH, Jones M, Kry S, Perry J, van de Kamer J, Verellen D, Jornet N. A commissioning protocol for portal imaging-based radiotherapy in vivo dosimetry systems. Phys Imaging Radiat Oncol 2024;32:100666. [Crossref] [PubMed]
12. Boggula R, Lorenz F, Mueller L, Birkner M, Wertz H, Stieler F, Steil V, Lohr F, Wenz F. Experimental validation of a commercial 3D dose verification system for intensity-modulated arc therapies. Phys Med Biol 2010;55:5619-33. [Crossref] [PubMed]
13. Venkataraman S, Malkoske KE, Jensen M, Nakonechny KD, Asuni G, McCurdy BM. The influence of a novel transmission detector on 6 MV x-ray beam characteristics. Phys Med Biol 2009;54:3173-83. [Crossref] [PubMed]
14. Defoor DL, Stathakis S, Roring JE, Kirby NA, Mavroidis P, Obeidat M, Papanikolaou N. Investigation of error detection capabilities of phantom, EPID and MLC log file based IMRT QA methods. J Appl Clin Med Phys 2017;18:172-9. [Crossref] [PubMed]
15. Dogan N, Mijnheer BJ, Padgett K, Nalichowski A, Wu C, Nyflot MJ, Olch AJ, Papanikolaou N, Shi J, Holmes SM, Moran J, Greer PB. AAPM Task Group Report 307: Use of EPIDs for Patient-Specific IMRT and VMAT QA. Med Phys 2023;50:e865-903. [Crossref] [PubMed]
16. Esposito M, Villaggi E, Bresciani S, Cilla S, Falco MD, Garibaldi C, Russo S, Talamonti C, Stasi M, Mancosu P. Estimating dose delivery accuracy in stereotactic body radiation therapy: A review of in-vivo measurement methods. Radiother Oncol 2020;149:158-67. [Crossref] [PubMed]
17. Li Y, Xiao F, Liu B, Qi M, Lu X, Cai J, Zhou L, Song T. Deep learning-based 3Din vivodose reconstruction with an electronic portal imaging device for magnetic resonance-linear accelerators: a proof of concept study. Phys Med Biol 2021;66. [Crossref] [PubMed]
18. Esposito M, Mancosu P, Bruschi A, Ghirelli A, Pini S, Alpi P, Barca R, Paoli CD, Meacci F, Leonulli BG, Fondelli S, Paoletti L, Scoccianti S, Russo S. The role of EPID in vivo dosimetry in the risk management of stereotactic lung treatments. Strahlenther Onkol 2023;199:992-9. [Crossref] [PubMed]
19. Baek TS, Chung EJ, Son J, Yoon M. Feasibility study on the verification of actual beam delivery in a treatment room using EPID transit dosimetry. Radiat Oncol 2014;9:273. [Crossref] [PubMed]
20. Ma C, Wang R, Zhou S, Wang M, Yue H, Zhang Y, Wu H. The structural similarity index for IMRT quality assurance: radiomics-based error classification. Med Phys 2021;48:80-93. [Crossref] [PubMed]
21. Wolfs CJA, Canters RAM, Verhaegen F. Identification of treatment error types for lung cancer patients using convolutional neural networks and EPID dosimetry. Radiother Oncol 2020;153:243-9. [Crossref] [PubMed]
22. Olaciregui-Ruiz I, Torres-Xirau I, Teuwen J, van der Heide UA, Mans A. A Deep Learning-based correction to EPID dosimetry for attenuation and scatter in the Unity MR-Linac system. Phys Med 2020;71:124-31. [Crossref] [PubMed]
23. Chytyk-Praznik K, VanUytven E, vanBeek TA, Greer PB, McCurdy BM. Model-based prediction of portal dose images during patient treatment. Med Phys 2013;40:031713. [Crossref] [PubMed]
24. Mahdavi SR, Tavakol A, Sanei M, Molana SH, Arbabi F, Rostami A, Barimani S. Use of artificial neural network for pretreatment verification of intensity modulation radiation therapy fields. Br J Radiol 2019;92:20190355. [Crossref] [PubMed]
25. Gimeno J, Pujades MC, García T, Carmona V, Lliso F, Palomo R, Candela-Juan C, Richart J, Perez-Calatayud J. Commissioning and initial experience with a commercial software for in vivo volumetric dosimetry. Phys Med 2014;30:954-9. [Crossref] [PubMed]
26. Kang S, Li J, Ma J, Zhang W, Liao X, Qing H, Tan T, Xin X, Tang B, Piermattei A, Orlandini LC. Evaluation of interfraction setup variations for postmastectomy radiation therapy using EPID-based in vivo dosimetry. J Appl Clin Med Phys 2019;20:43-52. [Crossref] [PubMed]
27. Ahmed S, Kapatoes J, Zhang G, Moros EG, Feygelman V. A hybrid volumetric dose verification method for single-isocenter multiple-target cranial SRS. J Appl Clin Med Phys 2018;19:651-8. [Crossref] [PubMed]
28. Zhuang AH, Olch AJ. Sensitivity study of an automated system for daily patient QA using EPID exit dose images. J Appl Clin Med Phys 2018;19:114-24. [Crossref] [PubMed]
29. Mijnheer B, Jomehzadeh A, González P, Olaciregui-Ruiz I, Rozendaal R, Shokrani P, Spreeuw H, Tielenburg R, Mans A. Error detection during VMAT delivery using EPID-based 3D transit dosimetry. Phys Med 2018;54:137-45. [Crossref] [PubMed]