H Sheen¹*, H Shin², H Kim³, C Kim⁴, J Kim⁵, J Kim⁶, C Hong⁷, (1) Heesoon Sheen1,*, Han-Back Shin2,*, Hojae Kim3, Changhwan Kim2, Jihun Kim2, Jin Sung Kim2, Chae-Seon Hong2,a) Author Institutions: 1Department of Health Sciences and Technology, Samsung Advanced Institute for Health Sciences & Technology, Sungkyunkwan University 2Department of Radiation Oncology, Yonsei Cancer Center, Yonsei University College of Medicine, Seoul, South Korea 3Department of Radiation Oncology, Yonsei Cancer Center, Seoul, South Korea

• 
  H Sheen

PO-GePV-T-199 (Sunday, 7/10/2022)   [Eastern Time (GMT-4)]

ePoster Forums

Purpose: This study aims to develop and evaluate the prediction model of multi-leaf collimator position (MLC) systematic errors using the using the index defined as error-related patterns.

Methods: AAPM TG-119 cases were planned according to guidelines, and then systematic errors were simulated. The gamma index, SSIM, and dosimetry index were used to analyze the dose distribution once without error and a second time with the simulation error. Spearman’s rank correlation of indices was investigated using the Holm-Bonferroni correction method, and the prediction model of MLC position systematic error was developed. The data were separated into training and test datasets. All models were built using repeated 10-fold cross-validation (CV) 10 times, and the finalized logistic regression models were evaluated using the area under the curve (AUC) of the receiver operating characteristic curve (ROC) in terms of sensitivity, specificity, accuracy, and precision.

Results: The effect of the MLC position systematic error on the dose volume histogram (DVH) was determined according to the location of the target and critical organs, complexity of the plan, and degree of error. Among the gamma index and SSIM index, there were no meaningful indexes, and only the dosiomics index was selected as a meaningful index. The GLRLM_LRHGE was selected for the analysis of MLC systematic error as a common dosiomics index. As the final models, 12 univariate predictive models and 3 multivariate models were developed. Their AUC was 0.8 or more (p < 0.05), and their accuracy, precision, sensitivity, and specificity was 0.8 or more (p < 0.05).

Conclusion: Our results clearly demonstrated the characteristics of MLC position systematic error and showed the robust performance of the error prediction model using relevant indices. The research results will be used as basic data to identify and study indicators applied to PSQA-related error analysis.

Funding Support, Disclosures, and Conflict of Interest: This work was supported by a National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (No. NRF-2021R1F1A1050903, NRF-2020R1C1C1005713, 2021R1F1A1050903).

Keywords

Quality Assurance, Radiation Therapy

Taxonomy

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

Contact Email