자기공명영상에서 표준화된 ACR 팬텀 영상을 통한 상용화된 딥러닝 기법 평가 = Evaluation of Deep Learning Techniques Commercialized in Magnetic Resonance Imaging through Standardized ACR Phantom Imaging

다국어 초록 (Multilingual Abstract)

Compared to existing imaging equipment, MR images exhibit a relatively large difference in reproducibility according to hydrogen density in the human body. It is therefore extremely important for the DL algorithm to demonstrate this in clinical practice and supplement it when a problem is found. Therefore, this study evaluated the changes before and after application of the DL algorithm by implementing the method now clinically applied to image quality standards using the MR phantom of the American College of Radiology (ACR) recommended by special medical equipment. The results revealed that aspects of the resolution-related image quality, such as high illuminance spatial resolution, improved resolution, and location accuracy were statistically different from the previous DL algorithm before application. There was no difference between images in the intensity of the DL algorithm. These results meet the evaluation criteria of ACR Phantom applied to the special medical equipment image quality management regulations, but the existence of statistical differences before and after application of the DL algorithm suggest other related research tools will be needed in the future.