A clinically observed discrepancy between image-based and log-based MLC positions

Purpose: To present a clinical case in which real-time intratreatment imaging identified an multileaf collimator (MLC) leaf to be consistently deviating from its programmed and logged position by >1 mm. Methods: An EPID-based exit-fluence dosimetry system designed to prevent gross delivery errors was used to capture cine during treatment images. The author serendipitously visually identified a suspected MLC leaf displacement that was not otherwise detected. The leaf position as recorded on the EPID images was measured and log-files were analyzed for the treatment in question, the prior day’s treatment, and for daily MLC test patterns acquired on those treatment days. Additional standard test patterns were used to quantify the leaf position. Results: Whereas the log-file reported no difference between planned and recorded positions, image-based measurements showed the leaf to be 1.3±0.1 mm medial from the planned position. This offset was confirmed with the test pattern irradiations. Conclusions: It has been clinically observed that log-file derived leaf positions can differ from their actual position by >1 mm, and therefore cannot be considered to be the actual leaf positions. This cautions the use of log-based methods for MLC or patient quality assurance without independent confirmation of log integrity. Frequent verification of MLC positions through independent means is a necessary precondition to trust log-file records. Intratreatment EPID imaging provides a method to capture departures from MLC planned positions. C 2016 American Association of Physicists in Medicine. [http://dx.doi.org/10.1118/1.4949002]


INTRODUCTION
While AAPM TG-142 recommends qualitative weekly and quantitative monthly quality assurance (QA) of multileaf collimator (MLC) leaf positions, 1 our clinic additionally performs a daily QA leaf position check.Until recently, our daily test required the therapist to judge if the light-field projected through a test pattern was within 1 mm of the pattern template.A recent clinical observation hastened our upgrade to quantitative image-based daily MLC position QA and exposed a discrepancy which cautions the use of MLC log-files for routine MLC or patient-specific QA.
The use of MLC logs for MLC and patient-specific QA stems from the belief that the log records display the actual leaf position at a time point during treatment. 2,3Retrospective studies which compared logged to planned positions for prior treatments found good agreement, thereby validating treatment delivery. 4Patient QA has been implemented using logged leaf positions to perform safety checks, 3 and commercial vendors have released log-file based patient QA products (MobiusFX, Mobius Medical, Houston, TX; Compass QuickCheck, Iba Dosimetry, Schwarzenbruck, Germany).
Imaging is a direct method for measuring the leaf position.While agreement between logged and EPID measured leaf positions was typically found, position variations of up to 1.5 mm for test fields have been reported. 2Based on this observation, Agnew et al. note that log-file based QA should not be solely relied upon as it does not always detect systematic machine faults. 2 In essence, they call into question whether logs represent actual leaf positions, which is a base assumption of log-file based QA.In this report, we provide further support for this question via a clinical observation recorded by our real-time EPID-based treatment monitoring software.During a patient's treatment, although planned and logged leaf positions were in close agreement, the during treatment imaging revealed a 1.3 ± 0.1 mm difference (at the isocenter plane) in one leaf position.

MATERIALS AND METHODS
As part of a developing real-time patient QA effort, we actively monitor patient exit-fluence using an aS1000 portal imager on a TrueBeam linear accelerator for some IMRT and VMAT treatments.Utilizing the WatchDog system, 5 images acquired at ∼8 Hz are compared to treatment-plan-based predictions with the aim of preventing gross treatment delivery error.Details of the WatchDog system are not presented here, as they have no bearing on the findings.
While performing real-time monitoring for a patient, a single leaf pair (Leaf pair 16) was visually observed by the first author to be displaced with respect to the neighboring leaf pairs, persisting throughout a VMAT treatment.An example frame is shown in Fig. 1.The variation was not observed in the previous day's treatment and was not autodetected by the software, but it cannot be considered a gross treatment error.
At the time of this observation, our facility's daily MLC QA consisted of comparing the light-field through a test pattern to a template.No discrepancies were noted by the therapist performing the test on the treatment days considered here.Fortuitously, a portal image of the test pattern was taken each day as a part of a real-time acquisition functionality test.
Postclinical observation image analysis of a test pattern image acquired prior to treatment showed that on the day of the observed displacement, while leaf 16A had a planned position (at isocenter plane) of 5.0 mm distal from midline, its actual position was 3.7 mm distal from midline [Fig.2(a)].(The leaf position was manually measured in imaging editing software.)The log-file recorded position during this acquisition was 4.989 mm, which was within 0.011 mm of the planned position, but 1.3 mm from the image-observed position.On the previous day, the test pattern image shows the leaf in the expected position, and the log-file reports a position of 4.993 mm.The absolute intensity difference between the portal images from the day of the displacement and the previous day is shown in Fig. 2(b).
QA with our standard test pattern following the observed during-treatment displacement showed a portal-image-based leaf displacement of 3.7 mm, a log-file-reported position of 4.989 mm, and just over 1 mm of visual displacement using the light-field and template.Picket-fence and sliding-window test pattern irradiations verified the out-of-tolerance condition.As such, the MLC was reinitialized in service mode.After reinitialization, the leaf was then observed to be at the intended position by film and light template, agreeing with a logged position of 5.004 mm.To confirm the test-pattern manual image-based displacement measurement, the test pattern was modified to have the leaf position at 3.5-3.8mm in 0.1 mm increments with images captured for each test pattern.An image with the leaf intentionally positioned at 3.7 mm showed the minimum difference in intensity when compared to the daily test pattern images with the misplaced leaf, hence we estimate the offset to be 1.3 ± 0.1 mm.
No other unusual activities or maintenance took place over this time frame.A service inspection postoccurrence found all equipment working properly and in proper condition.

DISCUSSION
Although our clinic had a daily visual MLC position check in place, it proved to be inadequate to detect the observed leaf offset.From quantitative image-based measurements, it is apparent that in this instance, the log-file misrepresents the true position, confirming the test-pattern findings of Agnew. 2 While logs provide precise leaf positions, routine calibration (e.g., via initialization) will ensure that the log reflects a more accurate position of the leaf in space. 6,7Litzenberg suggests semiweekly radiographic tests to identify position encoder drift.The sudden appearance of a large displacement such as this suggests that even the daily testing regimen used by Agnew 2 may allow single patient treatments over the ±0.5 mm MLC position tolerance recommended in the 2008 ESTRO guidelines. 8It seems apparent that real-time image-based QA may be a solution to this dilemma.
This finding has hastened our clinical change to daily quantitative image-based MLC position verification.It serves as a caution against using light-fields for daily MLC QA, and against using log files for MLC QA, patient QA, or patient dose verification.Leaf positions represented in the logs are not necessarily the actual leaf positions.

CONCLUSION
Based on our observation, log-file derived leaf positions cannot be considered to be the actual leaf positions without independent verification.Frequent (ideally continuous) independent verification of leaf position is necessary to ensure that treated MLC positions match planned positions.We encourage others who have found discrepancies between logged and actual positions to report findings (e.g., to an incident learning system), so we may understand the frequency and magnitude of such incidents.

F. 1 .
Real-time patient treatment images from consecutive days.The image on the right shows an MLC leaf displaced with respect to the neighboring leaves, and with respect to the image on the left from the previous day.This leaf offset was persistent throughout the beam delivery.

F. 2 .
(a) Image of the test pattern, showing leaf 16A displaced 1.3 mm.(b) Difference in image intensity between test pattern images on the day of the anomaly and the previous day.The largest difference visible at the top of the MLC diamond pattern results from a single leaf out of position.The rectangular-shaped images are the result of small displacements in an imaging phantom, unrelated to the MLC.Medical Physics, Vol.43, No. 6, June 2016