H Mehrens*, N Hernandez, B Lewis, A Molineu, S Kry, The University of Texas MD Anderson Cancer Center, Houston, TX

• 
  H Mehrens

• 
  
• 
  
• 
  
• 
  
• 
  
• 
  
• 
  

SU-F-BRA-1 (Sunday, 7/10/2022) 2:00 PM - 3:00 PM [Eastern Time (GMT-4)]

Ballroom A

Purpose: To analyze institutional performance irradiating IROC Houston’s H&N IMRT phantom to identify traits that are most predictive of a failing phantom result.

Methods: Irradiation results of 1542 HN phantoms between 2012-2020 were abstracted. Univariate analysis, factor analysis, clustering, and random forest models were used to associate irradiation conditions and plan complexity scores with phantom results. Phantom results included pass/fail classification, average TLD ratio for OAR and two primary targets, and percent of pixels passing gamma. The following treatment parameters were evaluated in terms of how they predicted these outcomes: planning system, irradiation year, algorithm, machine model, energy, irradiation technique, and the following complexity scores: mean tongue and groove index (meanTGi), MLC speed modulation, modulation complexity score (MCS), total modulation index (MIt), plan irregularity, edge metric (EM), leaf travel, and first quartile of MLC gap sizes (Q1Gap).

Results: All complexity scores, except MCS, indicated that plans have increased in complexity (p<0.05; regression) over the data’s timeframe. In terms of predicting pass/fail, AAA was statistically superior to pencil beam (p = 0.004; chi squared) and superposition convolution (S/C) algorithms (p = 0.001). In addition, XiO planning system, segmental irradiation technique, and Elekta and Siemens machines were statistically inferior compared to most other constituents within their category. For overall pass/fail, TLD results, and percent of pixels passing gamma, Q1Gap was the most important predictor. The next three most important predictors were meanTGi, EM, and MCS; however, the order of importance varied depending on which endpoint was being predicted.

Conclusion: Several treatment planning parameters, such as, pencil beam and S/C algorithms, performed poorly compared to other constituents within their category. Important predictive complexity scores may elucidate how these treatment parameters can improve pass rates. With complexity increasing in plans, great care must be taken when utilizing such systems to ensure proper dose delivery.

Funding Support, Disclosures, and Conflict of Interest: This work was supported by grants CA214526 and CA180803 awarded by the NCI.

Keywords

Quality Assurance, Phantoms

Taxonomy

TH- External Beam- Photons: Quality Assurance - IMRT/VMAT

Contact Email