Real-time monitoring of linac performance using RT plans and...
Transcript of Real-time monitoring of linac performance using RT plans and...
Real-time monitoring of linac performance using RT plans and logfiles
NCCAAPM – Wisconsin Dells7 April 2017
Mark Wiesmeyer, PhD, DABRTechnical Product Manager
Technical talk with insights and results based on a commercial product.
Goals
• Check that a machine can deliver plans that you create*.
• Get real‐time feedback about plan delivery issues**.
• Create a longitudinal record of plan delivery quality.
• Determine if a machine may need mechanical calibration.
* Pre‐treatment IMRT QA** During treatment QA
I hope that you will gain a further understanding of DICOM RT Plans and Logfiles.
Roadmap• Testable Parameters
• Basic Information Flow
• RT Plans – Control Points
• Logfiles – Sampled Machine Data Records
• Mapping Logfile Records to RT Plan Control Points
• Some results
Testable Parameters
• Jaw Positions
• Carriage Positions
• MLC Leaf Positions
• Gantry and Collimator Angles
• Fraction of total MU
• Beam States (Hold, On/Off, etc.)
• Planned vs Delivered Fluence
• Table Positions and Angles
Basic Information Flow
Treatment Planning Computer
Linac Console
Logfiles
RT Plans RT PlansLogfile Batch Folder
RT Plan Batch Folder
LinacView Computer
1 3
2
RT Plans – General Abridged
Where the “action” is…
RT Plans – Beam Sequence
Where the “action” is…
RT Plan - Control Points (CP1)
Gantry AngleCollimator Angle
Couch Angle
X Jaws or Carriages
Y Jaws or Carriages
MLC Positions
Fractional MU’s
Control Points (CPs 88,89,178)
Fractional MU’sGantry Angle
Fractional MU’sGantry Angle
Fractional MU’sGantry Angle
Quantities whose values do not change are not repeated in CP’s.
Common Systems that Provide Logfiles
• Sampling rate – the number of data records collected per second.
• Faster sampling is typically better.
• 4 Hz really only suitable for static analysis.
• Contents of DynaLog, Trajectory, and TRF files are comparable.
• Key problem and source of error is mapping logfile records to CP’s.
Vendor Linacs Logfile Format Sampling (Hz) CommentsVarian Clinic Series DynaLog 20 CSV (spec).
TrueBeams Trajectory 50 Binary (spec).Elekta All Recent Linacs *.dat, *.xml 4 iCom Vx API (spec).
Agility 160 Leaf TRF Logfiles 25 Binary (no spec).
Sample Logfile Specification – DynaLog
DynaLog Files
Originally intended for maintenance QA…
1 – 25000 mapping
Notice: No explicit mapping to control points.
Dose Fraction:150 MU(Rx) * 2198/25000 = 13.2 MU.
50 ms/row
Trajectory Files
Much easier to match to RT Plan CP’s…
Notice: Direct, fractional control point mapping.Direct, fractional MU’s.
20 ms/row
DynaLog File Mapping
RT‐Plan Control Points precede Logfile Samples
• Linear interpolation maps measurement records to control points.
• Worst case uncertainty is 25 ms…”Registration Error”.
• For leaves this results in about: 2.5 cm/s * 0.025 s = 625mm additional error
Max Leaf Speed.
Trajectory File Mapping
• Measurement records to map directly to fractional control points.
• Worst case uncertainty / registration error is negligible.
Clinac/DynaLog (Name 04)
Fluence maps are created by projecting MU weighted segments to the isocenter plane.
Gamma index results – 99.7%
RT Plan Fluence Map
Logfile Fluence Map
TrueBeam/Trajectory(Name 18)
RT Plan Fluence Map
Logfile Fluence Map
Gamma index results – 100%
Good results are not surprising. • Compare simulated fluences not measurements.• Comparison is as fine as needed.• No finite‐spaced detector grids.• No setup uncertainty.
Leaf Positions & MU’s Clinac/DynaLog
In addition to “registration error”, actual errors from MLC position may be caused by:
• Faulty or loose T‐nuts• Motor failures • Encoder drift• Dirty, sticky leaves
MU delivery error range:‐0.10 ‐> 0.15 MU
How much is Registration Error?
Worst case per leaf error for this beam delivery.
Leaf Positions & MU’s TrueBeam/Trajectory
Little or no “registration error” and a well‐maintained, somewhat more precise machine.
MU delivery error range:‐0.01 ‐> 0.01 MU
What can Logfile analysis tell us?A. Your plan will deliver / is delivering correct dose to your patient.
B. Your machine is capable of delivering the plan that you have created.
C. Your machine is going out of mechanical calibration. Not dose
D. All of the above.
E. A & B only.
F. B & C only.
For dose tracking, patient setup and geometry changes are at least an order of magnitude larger than machine delivery errors which tend to be very small.