Hausdorff Distance Between Plan and Delivery Logs for Daily Treatment Quality Assurance
Y Hao*, B Sun, A Darafsheh, T Mazur, W Smith, T Zhang, T Zhao, Washington University School of Medicine, St. Louis, MO
Presentations
TU-IePD-TRACK 5-6 (Tuesday, 7/27/2021) 3:00 PM - 3:30 PM [Eastern Time (GMT-4)]
Purpose: Intensity-modulated proton therapy (IMPT) provides better treatment quality than conventional radiotherapy at the cost of increased complexity. Parameters that could cause catastrophic consequences include energy, spot location, spot weight (MU), and leaf positions for some proton units. This work proposes a novel method for quality assurance (QA) of daily IMPT treatment records.
Methods: This method was tested on a synchrocyclotron unit equipped with adaptive apertures for shaper penumbra with layer-by-layer and spot-by-spot collimation. Four types of parameters were extracted from daily delivery logs and arrange into a 35-D Hausdorff space (HS), including 18 range shifter plates for energy layer, 14 leaves (7 pairs) for the adaptive aperture, 2 coordinates for spot position, and weight. A unitless weighted Hausdorff Distance (HD) was calculated between the planned and delivered HS for all pulses in a field. The weight of each dimension in the HS is based on the severity of any deviation. An automated report tool was developed to verify treatment records for all patients daily.
Results: The tool reported deviations in delivery with HD larger than 5. Among the 42 fractions tagged by the tool in the last two months, 36 were not reported by our R&V system. Those errors have been analyzed, and none caused significant dosimetric concern. The analysis provides valuable information identifying the spot index, energy layer, and range shift plates involved in any fault to support repair and preventative maintenance.
Conclusion: Our machine log file analysis can catch errors originating from data transfer problems, machine delivery issues, and other human mistakes. The dosimetric impact of these errors can be recomputed based on machine delivery conditions. The historic failure trends can assist in maintaining system operation at high efficiency.
ePosters
Keywords
Not Applicable / None Entered.
Taxonomy
Contact Email