TL;DR
This paper investigates the significant variability in monitor units (MU) observed in online adaptive radiotherapy (oART) treatments using the Ethos system. They found that these MU variations are not due to errors in the workflow but rather inherent to the optimizer, and do not correlate with anatomical changes. This suggests that clinicians should rely on secondary dose calculations, not MU fluctuations, to ensure treatment accuracy.
What Problem Does This Paper Address?
Online adaptive radiotherapy (oART) aims to improve treatment accuracy by adjusting radiation plans daily based on patient anatomy changes detected via Cone-Beam CT (CBCT). While this technology promises better outcomes, the rapid nature of on-couch adaptation presents challenges for clinical teams in verifying plan quality. Specifically, the authors observed substantial day-to-day variations in the total monitor units (MU) required for treatment delivery, sometimes exceeding 30% compared to the initial plan. These large MU discrepancies raised concerns about the accuracy of the dose calculation within the oART system and prompted this investigation into the causes and implications of this MU variability.
Methods
The researchers analyzed dosimetric and adaptive plan data from 604 sessions across 19 patients undergoing CBCT-guided oART. Their approach involved:
- Data Collection: Gathering total MU per fraction, PTV and OAR volumes, and DVH curves from the Ethos system.
- Statistical Analysis: Evaluating the distribution of MU variations, PTV/OAR volume differences for normality using the Shapiro-Wilk test.
- Correlation Analysis: Investigating potential correlations between MU variations and changes in PTV volume or OAR-PTV overlap using Pearson correlation coefficients.
- Optimizer Investigation: Reoptimizing selected fractions with significant MU variations (greater than 2 standard deviations) using the Ethos Treatment Planning Management System (TPMS) to assess reproducibility and compare it to the on-couch optimizer.
- Independent Verification: Performing secondary dose calculations using Mobius3D and phantom measurements (ArcCheck and ion chamber) for fractions with substantial MU variations to confirm the accuracy of the delivered dose.
Datasets
The study utilized a dataset of 604 on-couch adaptive sessions from 19 patients treated with the Ethos system. This data was retrospectively gathered from the Mobius3D database. No new benchmarks were introduced; the study focused on analyzing existing clinical data.
Key Insights and Ideas
- Inherent Optimizer Variability: The primary finding is that MU variations in oART are not indicative of workflow errors but are likely a consequence of the stochastic nature of the Intelligent Optimization Engine (IOE) within the Ethos system.
- Anatomy Independence: Crucially, MU variations did not correlate with changes in patient anatomy (PTV or OAR volumes, or their overlap). This suggests that the optimizer is adjusting MU for reasons unrelated to the actual anatomical shifts.
- TPMS vs. On-Couch Optimizer: Reoptimizing plans in the TPMS showed similar variability to the on-couch optimizer, reinforcing the idea that the variability is a characteristic of the underlying optimization algorithms rather than a specific implementation issue.
- Dose Calculation Accuracy is Reliable: Despite the MU fluctuations, independent dose calculations and phantom measurements confirmed that the dose delivered by the Ethos system is accurate and within acceptable clinical tolerances.
Key Results
- MU variations relative to the reference plan followed a normal distribution with a mean of -1.0% and a standard deviation of 11.0%.
- Fourteen fractions (22%) exhibited MU variations greater than 2 standard deviations from the mean.
- No significant correlation was found between MU variation and changes in PTV volume, OAR volume, or PTV-OAR overlap.
- Offline reoptimization in the TPMS did not reliably reproduce reference or on-couch total MUs, indicating similar variability.
- Independent dose calculations and detector array measurements showed acceptable agreement with the planned dose, with gamma passing rates consistently high.
Limitations and Future Work
- Generalizability: The study was conducted at a single institution using the Ethos system. Further validation across different institutions and oART platforms would be beneficial.
- Anatomical Correlation Nuance: While no significant correlation was found, the paper mentions that MU variations can be influenced by anatomical changes, but the magnitude of these variations was not reliably reproducible. This suggests a complex interplay that warrants further investigation.
- Future Work: The authors suggest that clinical decisions during oART should rely on validated secondary dose calculations rather than MU variability to ensure optimal plan selection. Future work could explore methods to better predict or control MU variability, or to more effectively integrate MU information into the clinical decision-making process without misinterpreting it as an indicator of accuracy.
Summary updated