We konden het bij MST in eerste instantie ook niet geloven, erkent bestralingsdeskundige Liselotte ten Asbroek. Toen de afdeling radiotherapie zo’n drie jaar geleden op het punt stond om een van de bestralingsapparaten te vervangen, kwam het uit bij Varian. De fabrikant van apparatuur en software voor radiotherapie beloofde met haar nieuwe bestralingssysteem Ethos Therapy zowel de voorbereiding als de duur van een radiotherapiebehandeling sterk te kunnen reduceren.
Voorafgaand aan een behandeling wordt een radiotherapieplan opgesteld. Dit plan moet dagelijks worden aangepast aan veranderingen in de anatomie. Ten Asbroek: “Om het plan aan te passen, waren we in de oude situatie een dag bezig, en dan kon de anatomie van de patiënt alweer veranderd zijn. Dat dit aanpassen nu bijna meteen kan (in ongeveer een kwartier), is iets dat we ons toen niet konden voorstellen.” Afdelingsmanager Radiotherapie André Doldersum: “De haarscherpe beeldkwaliteit van de Cone Beam CT-scan van de Ethos en de kunstmatige intelligentie (AI) in de software eromheen maken dit mogelijk. Hiermee is de deur opengezet voor een nieuwe aanpak van radiotherapie.”>>>>>>>lees verder
Je zou het de nieuwste Ferrari of Porsche onder de tumorenbestralers kunnen noemen: de Varian Ethos. Hij is supersnel en heel nauwkeurig. MST is het enige ziekenhuis in Nederland dat er eentje bezit.
De Ethos is het allernieuwste dat er op het gebied van bestraling te koop is én het allerbeste. De voordelen van de machine liggen vooral bij de patiënt, want hij is niet alleen snel, hij is ook uiterst nauwkeurig. Dat betekent: gelijk raak schieten en dus minder klachten. Voorlopig worden met de Varian Ethos Therapy alleen prostaatkankers bestraald, maar in de toekomst wordt hij ook voor andere tumoren ingezet.
De Ethos ziet eruit als een grote ring, waarin de patiënt geschoven wordt op een tafel. Technisch gezien is de Ethos een deeltjesversneller die fotonen op de tumor afvuurt. Daar kan het dna niet tegen en zo wordt de tumor gedood. Dat klinkt eenvoudiger dan het is, want in het lichaam is het nooit rustig. Dat maakt het ook lastig om precies te mikken, aldus radiotherapeut en oncoloog Dankert Woutersen van MST. „Een volle endeldarm, een gevulde blaas of darmgas kunnen de organen een beetje van de plek duwen.”
Zijn collega Liesbeth de Wit weet daar alles van. De radiotherapeut en oncoloog is gespecialiseerd in het bestralen van prostaten. „De prostaat kan soms net op een iets andere plek zitten”, zegt ze. „Vroeger maakten we één behandelplan en dan bestraalden we steeds dezelfde plek. Toen wisten ook al wel dat er wat beweging in de prostaat kan zitten. Dus moesten we aan de veilige kant blijven.”
The frequency of minor edits, or no edits, to influencer and target contours found in this study overall was 92% and 91%, respectively. This is broadly in agreement with that shown by Sibolt et al. 18 who reported 76% of influencers requiring no or minor edits over a greater range of treatment types. The results of Sibolt et al. are also somewhat skewed by the inclusion of bladder cancer treatments with a catheter in place, which were noted to have performed poorly during AI contouring. The frequency of target edits is marginally higher than that seen in the study by Sibolt et al. 18 for prostate (100% requiring no or minor changes), but that study did not include prostate bed cases, or separate nodal from primary CTVs in the data presented. The results show that the intact prostate CTVp required less edits than the prostate bed CTVp, indicating that the system performs more accurately where there is a GTV that is also an influencer, compared to the more variable prostate bed CTVs without a GTV. Even so, the reported frequency of editing of influencer and target contours was similar, and indicates an efficient workflow is possible. The previous study by Yoon et al. 19 covered different anatomy, however the reported accuracy of contouring was broadly consistent with this study. The frequency of adaptive plan selection reported by Sibolt et al. was lower than found in this study (88% compared with 95% found here), although not inconsistent given the range seen for different treatment sites (89.4% for prostate and nodes).
The differences in contouring accuracy for the sigmoid colon seen in Figure in prostate bed cases are thought at least partially to be due to more mobility in the sigmoid colon post‐prostatectomy. Due to the propagation of the sigmoid colon from the simulation CT to the CBCT using an elastic DIR (which is not affected by influencers) in an area of low HU contrast, the DIR poorly tracks sigmoid colon movement. Therefore, the daily sigmoid colon contour accuracy is highly dependent on the sigmoid position during the simulation CT, and specifically how representative it is of the average sigmoid colon position on the treatment. The data presented here include 125 fractions, but these correspond to just six simulation CT scans, and therefore are likely an insufficient sample size to predict frequency of sigmoid colon edits.
When edits are made to targets or noninfluencer OARs, the treatment time substantially lengthens, as the optimization process that is already underway during the contour review step is restarted. Therefore to maximize efficiency, ideally no edits are required, and even minor changes can be problematic if seen consistently. To avoid the frequent edits to the sigmoid colon, we propose to add a 5 mm sigmoid PRV to the workflow in future. This PRV structure would be used as a dose avoidance region and would not be regenerated daily based on the sigmoid colon, but rather propagated from the planning CT. This would allow the user to verify that the sigmoid is within the PRV daily, and thus not spend additional time editing the sigmoid colon which would force a re‐optimization of the plan. This alteration has the disadvantage that it is not adapting the plan to the daily position of the sigmoid colon but rather is avoiding dose in the whole region surrounding the sigmoid colon, and thus may unnecessarily compromise coverage. In future, an AI‐based sigmoid colon influencer would be a preferable solution.
The contouring results for the single patient with a hip prosthesis were poor and required extensive editing. While the beam arrangement used can be adjusted to avoid the prosthesis, the extended contouring time is likely to make the use of Ethos adaptive inadvisable for the treatment of patients with hip prostheses.
This study employed user‐reported frequency and magnitude of contour editing. It is acknowledged that this is an imprecise surrogate for automated contouring accuracy. A more rigorous method would be to use a quantitative analysis (e.g., Dice similarity coefficient) comparing the automatically generated contours without edits against contours generated by an expert user (or users). Apart from being prohibitively cumbersome for a large number of fractions in the current implementation of Ethos, this would only allow analysis of OARs (both influencers and noninfluencers), as the CTVs are generated based on the edited influencers. One of the reasons that edits were required for the CTVs less frequently than for influencers, is that errors in contouring had been fixed in the influencer step and those fixes had been propagated to the CTVs. This is the intended workflow and design of software, so to analyze CTV accuracy without influencer edits would overestimate the CTV errors that occur in the actual treatment workflow. The other major reason that no changes are required for CTVs more often than no changes for influencers is the different instructions given to users, specifically not to make changes to CTVs unless the change is clinically significant.
In a small number of fractions, it was noted that the scheduled plan had more goals met than the adapted plan. Further analysis of these cases indicated that this could be due to some of the goals becoming more contradictory in a particular fraction (e.g., a rectum maximum dose overlapping with a minimum dose to the PTV). Due to the hierarchical nature of the IOE used in Ethos, this can mean that goals below this contradiction are not fully optimized, and lead to the scheduled plan meeting more clinical goals than the adaptive plan.
The largest plan quality differences noted for any structure was for the PTV, which consistently showed significant improvements in the adaptive plan, also seen in other studies. 26 This is anticipated, as the PTV is just a tool to ensure that the CTV is covered, and under a normal image‐guided treatment it is not expected that the PTV would be fully covered on any given fraction, so long as the CTV is covered. Differences in CTV coverage were much smaller between scheduled and adapted plans, which is indicative that the margins were suitable for nonadaptive treatments. This is expected as the margins were those previously used for IGRT and chosen to cover the CTV with expected setup uncertainties. Re‐evaluation of margins will be the subject of a future study and may lead to greater dosimetric improvements being observed. The large differences in plan quality reported in previous studies such as Ahunbay et al. 1 were primarily achieved through margin reductions, and therefore were not seen here.
Other plan differences were more mixed, with the adaptive plan sometimes inferior to the scheduled plan for a given clinical goal. When this occurred, it was generally noted that the adaptive plan had already met the ideal planning goal. Note that with the exception of the PTV discussed above, Table does not suggest that the statistically significant differences seen for the case shown will be replicated for all prostate adaptive cases. Rather, that the goals and priorities selected for this specific case led to statistically significant improvements and deteriorations in the areas shown. A different set of priorities and goals would likely lead to different statistically significant differences for other cases. Our study suggests that organs that change significantly day‐to‐day (such that a goal is exceeded) are likely to see the largest improvement with adaptive replanning. Additionally, the statistically significant results shown in Table do not necessarily indicate that the differences are also clinically significant.
The fraction timing data are consistent with what has been reported by other studies. 18 , 19 As would be expected, sites with more structures and more frequent contour editing took longer. It is expected the fraction time will reduce as staff continue to gain experience with the system.
Significant improvements in dosimetry are possible using the adapted plan, when compared to the scheduled plan, with the Ethos OART system for intact and prostate bed radiotherapy. No change or minor edits were achieved in 92% of influencer contours and 91% of target contours. The adaptive plan was selected in 95% of fractions. The early data presented here will assist other users of the Ethos system in implementing online adaptive radiotherapy to the prostate.
Mikel Byrne, Ben Archibald‐Heeren, Amy Teh, and Rhea Beserminji designed the study. Mikel Byrne, Ben Archibald‐Heeren, Yunfei Hu, Amy Teh, Rhea Beserminji, Emma Cai, Guilin Liu, and Angela Yates all participated in acquiring the data for the study, either by performing retrospective or clinical treatment fractions, or both. Data analysis was performed by Mikel Byrne, Ben Archibald‐Heeren, Yunfei Hu, James Rijken, Nick Collett, and Trent Aland. All contributing authors reviewed the manuscript and gave feedback on the findings.
CONFLICT OF INTEREST
Icon group is a member of the Varian Adaptive Intelligence Consortium and has a partnership with Varian to provide radiotherapy equipment. Mikel Byrne, Ben Archibald‐Heeren and Amy Teh have also received honoraria for presenting on behalf of Varian Medical Systems.
The authors would like to thank the Icon group for supporting this research. We would also like to thank all Icon Wahroonga staff that assisted with data acquisition for this project.
Byrne M, Archibald‐Heeren B, Hu Y, et al. Varian ethos online adaptive radiotherapy for prostate cancer: Early results of contouring accuracy, treatment plan quality, and treatment time. J Appl Clin Med Phys. 2022;23:e13479. 10.1002/acm2.13479 [PMC free article] [PubMed] [CrossRef] [Google Scholar]
1. Ahunbay EE, Peng C, Holmes S, et al. Online adaptive replanning method for prostate radiotherapy. Int J Radiat Oncol Biol Phys. 2010;77:1561‐1572. [PubMed] [Google Scholar]
2. Court LE, Dong L, Lee AK, et al. An automatic CT‐guided adaptive radiation therapy technique by online modification of multileaf collimator leaf positions for prostate cancer. Int J Radiat Oncol Biol Phys. 2005;62:154‐163. [PubMed] [Google Scholar]
3. Crijns W, Van Herck H, Defraene G, et al. Dosimetric adaptive IMRT driven by fiducial points. Med Phys. 2014;41:061716. [PubMed] [Google Scholar]
4. Hwang AB, Chen J, Nguyen TB, et al. Irradiation of the prostate and pelvic lymph nodes with an adaptive algorithm. Med Phys. 2012;39:1119‐1124. [PubMed] [Google Scholar]
5. Li X, Quan EM, Li Y, et al. A fully automated method for CT‐on‐rails‐guided online adaptive planning for prostate cancer intensity modulated radiation therapy. Int J Radiat Oncol Biol Phys. 2013;86:835‐841. [PubMed] [Google Scholar]
6. Liu H, Wu Q. Evaluations of an adaptive planning technique incorporating dose feedback in image‐guided radiotherapy of prostate cancer. Med Phys. 2011;38:6362‐6370. [PMC free article] [PubMed] [Google Scholar]
7. Schulze D, Liang J, Yan D, Zhang T. Comparison of various online IGRT strategies: the benefits of online treatment plan re‐optimization. Radiother Oncol. 2009;90:367‐376. [PMC free article] [PubMed] [Google Scholar]
8. Thongphiew D, Wu QJ, Lee WR, et al. Comparison of online IGRT techniques for prostate IMRT treatment: adaptive vs. repositioning correction. Med Phys. 2009;36:1651‐1662. [PubMed] [Google Scholar]
9. Nijkamp J, Pos FJ, Nuver TT, et al. Adaptive radiotherapy for prostate cancer using kilovoltage cone‐beam computed tomography: first clinical results. Int J Radiat Oncol Biol Phys. 2008;70:75‐82. [PubMed] [Google Scholar]
10. Tetar SU, Bruynzeel AME, Lagerwaard FJ, et al. Clinical implementation of magnetic resonance imaging‐guided adaptive radiotherapy for localized prostate cancer. Phys Imaging Radiat Oncol. 2019;9:69‐76. [PMC free article] [PubMed] [Google Scholar]
11. Mangar S, Coffey J, Mcnair H, et al. Prostate radiotherapy: evaluating the effect of bladder and rectal changes on prostate movement: a CT study. Trends Med Res. 2006;1:55‐65. [Google Scholar]
12. Holden L, Stanford J, D'Alimonte L, et al. Timing variability of bladder volumes in men receiving radiotherapy to the prostate. J Med Imaging Radiat Sci. 2014;45:24‐30. [PubMed] [Google Scholar]
13. Jarrett D, Stride E, Vallis K, Gooding MJ. Applications and limitations of machine learning in radiation oncology. Br J Radiol. 2019;92:20190001. [PMC free article] [PubMed] [Google Scholar]
14. Netherton TJ, Cardenas CE, Rhee DJ, et al. The emergence of artificial intelligence within radiation oncology treatment planning. Oncology. 2021;99:124‐134. [PubMed] [Google Scholar]
16. Spalding M, Walsh A, Aland T. Evaluation of a new GPU‐enabled VMAT multi‐criteria optimisation plan generation algorithm. Med Dosim. 2020;45:368‐373. [PubMed] [Google Scholar]
17. Aland T, Walsh A, Jones M, et al. Accuracy and efficiency of graphics processing unit (GPU) based Acuros XB dose calculation within the Varian Eclipse treatment planning system. Med Dosim. 2019;44:219‐225. [PubMed] [Google Scholar]
18. Sibolt P, Andersson LM, Calmels L, et al. Clinical implementation of artificial intelligence‐driven cone‐beam computed tomography‐guided online adaptive radiotherapy in the pelvic region. Phys Imaging Radiat Oncol. 2021;17:1‐7. [PMC free article] [PubMed] [Google Scholar]
19. Yoon SW, Lin H, Alonso‐Basanta M, et al. Initial evaluation of a novel cone‐beam CT‐based semi‐automated online adaptive radiotherapy system for head and neck cancer treatment—a timing and automation quality study. Cureus. 2020;12:e9660. [PMC free article] [PubMed] [Google Scholar]
20. Archambault Y, Boylan C, Bullock D. Making on‐line adaptive radiotherapy possible using artificial intelligence and machine learning for efficient daily re‐planning. Med Phys Int J. 2020;8:77‐86. [Google Scholar]
21. Hu Y, Byrne M, Archibald‐Heeren B, et al. Validation of the preconfigured Varian Ethos Acuros XB beam model for treatment planning dose calculations: a dosimetric study. J Appl Clin Med Phys. 2020;21:27‐42. [PMC free article] [PubMed] [Google Scholar]
22. Varian Medical Systems, Ethos Algorithms Reference Guide 2019.
23. McDonald DG, Jacqmin DJ, Mart CJ, et al. Validation of a modern second‐check dosimetry system using a novel verification phantom. J Appl Clin Med Phys. 2017;18:170‐177. [PMC free article] [PubMed] [Google Scholar]
24. Varian Medical Systems, Ethos Treatment Planning Instructions for Use 2019.
26. De Jong R, Visser J, Van Wieringen N, et al. Feasibility of conebeam CT‐based online adaptive radiotherapy for neoadjuvant treatment of rectal cancer. Radiat Oncol. 2021;16:136. [PMC free article] [PubMed] [Google Scholar]
Plaats een reactie ...
Reageer op "Varian Ethos is naast de MRiDian een nieuwe precisie bestralingsmachine die voorlopig alleen nog bij prostaatkankerpatienten wordt gebruikt"