Click here to


Are you sure ?

Yes, do it No, cancel

A Simple Statistical Method to Reduce the Number of Patient-Specific QA Measurements for MR-Linac Adaptive Fractions

A Kim1*, M Ruschin1, C Mccann1, P Au1, A Singh1, A Sahgal2, B Keller1, (1) Sunnybrook Odette Cancer Centre, Department of Medical Physics, Toronto, ON, CA, (2) Sunnybrook Health Sciences Centre, Odette Cancer Centre, Department of Radiation Oncology, CA


(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: To demonstrate a method to reduce the number of MR-Linac patient-specific QA measurements.

Methods: When we began treating with the MR-Linac in mid-2019, we performed patient-specific QA measurements with a commercial helical diode array phantom for every adapted fraction, since the plan-of-the-day is unique. QA passes with a gamma pass rate>95% with the plan/measurement dose distributions matching within 3%/3mm. We realized that measuring every fraction was an impractically high workload (e.g. most brain patients are treated with 30 fractions). Hence, we now use a statistical tool—the confidence interval for the mean—to justify measuring only the reference plan and the first adapted fraction.

A confidence interval is a range of values that we are reasonably sure our true mean is in. With patient-specific QA measurements, it is satisfactory if the mean of all of the gamma pass rates in a course is >95%. To use this method, we needed a representative standard deviation for brain patients. This data comes from our first five brain patients (81 measurements). For each patient the standard deviation of the gamma was calculated. The confidence interval was used to predict from few measurements if all the gamma results in the course will pass QA.

Results: For the five patients, the gamma standard deviations ranged from 0.0%-0.6%. 0.6% was selected as the worst case value for the confidence interval calculator. Entering in each of the first two measurements for each of the five patients, the calculator returns 99.9% confidence intervals. The true mean for all patients always lay within the confidence intervals.

Conclusion: We have demonstrated a method to calculate the confidence interval of gamma pass rates for an MR-Linac course, relying upon the first two measurements. This saves a large amount of work, as it statistically justifies not performing patient-specific QA on every fraction.


MRI, Quality Assurance, Performance Tests


IM/TH- MRI in Radiation Therapy: MRI/Linear accelerator combined Quality Assurance

Contact Email