Clinical Efficacy of an Electronic Portal Imaging Device versus a Physical Phantom Tool for Patient-Specific Quality Assurance

Clinical Efficacy of an Electronic Portal Imaging Device versus a Physical Phantom Tool for Patient-Specific Quality Assurance

Seung-Hyeop Baek^1,2,3 , Sang-Hyoun Choi^2,4 , Moo-Jae Han², Gyu-Seok Cho², Wonil Jang⁴, Jin-Sung Kim^3,5,6 and Kum-Bae Kim^2,4

 1 Department of Integrative Medicine, Yonsei University College of Medicine, Seoul 03722, Republic of Korea
2 Research Team of Radiological Physics & Engineering, Korea Institute of Radiological & Medical Sciences, Seoul 01812, Republic of Korea
3 Medical Physics and Biomedical Engineering Lab (MPBEL), Yonsei University College of Medicine, Seoul 03722, Republic of Korea
4 Department of Radiation Oncology, Korea Institute of Radiological & Medical Sciences, Seoul 01812, Republic of Korea
5 Department of Radiation Oncology, Yonsei Cancer Center, Heavy Ion Therapy Research Institute, Yonsei University College of Medicine, Seoul 03722, Republic of Korea
6 Oncosoft Inc., Seoul 03787, Republic of Korea

ABSTRACT

Pre-treatment patient-specific quality assurance (QA) is critical to prevent radiation acci-dents. The electronic portal imaging device (EPID) is a dose measurement tool with good resolution and a low volume-averaging effect. EPIbeam—an EPID-based portal dosimetry software—has been newly installed in three institutions in Korea. This study evaluated the efficacy of the EPID-based patient-specific QA tool versus the PTW729 detector (a previously used QA tool) based on gamma criteria and planning target volume (PTV). A significant difference was confirmed through the R statistical analysis software. The average gamma passing rates of PTW729 and EPIbeam were 98.73%and 99.60% on 3 mm/3% (local), 96.66% and 97.91% on 2 mm/2% (local), and 88.41% and 74.87%on 1 mm/1% (local), respectively. The p-values between them were 0.015 (3 mm/3%, local), 0.084 (2 mm/2%, local), and less than 0.01 (1 mm/1%, local). Further, the average gamma passing rates of PTW 729 and EPIbeam according to PTV size were 99.55% and 99.91% (PTV < 150 cm3) and 97.91%and 99.28% (PTV > 150 cm3), respectively. The p-values between them were 0.087 (PTV < 150 cm3) and 0.036 (PTV > 150 cm3). These results confirm that EPIbeam can be an effective patient-specific QA tool.Translation of predictive and prognostic image-based learning models to clinical applications is challenging due in part to their lack of interpretability. Some deep-learning-based methods provide information about the regions driving the model output. Yet, due to the high-level abstraction of deep features, these methods do not completely solve the interpretation challenge. In addition, low sample size cohorts can lead to instabilities and suboptimal convergence for models involving a large number of parameters such as convolutional neural networks.