Noise Level and Blind Quality Evaluations in Abdominopelvic Electromagnetic X-ray Computed Tomographic Images using Deep-learning Reconstruction Method
- Authors
- Shim, Jina; Lee, Seungwan; Lee, Youngjin
- Issue Date
- Mar-2024
- Publisher
- KOREAN MAGNETICS SOC
- Keywords
- Abdominopelvic computed tomography (CT); Deep-learning image reconstruction (DLIR) intensity; Noise level evaluation; Blind image quality evaluation
- Citation
- JOURNAL OF MAGNETICS, v.29, no.1, pp 63 - 70
- Pages
- 8
- Journal Title
- JOURNAL OF MAGNETICS
- Volume
- 29
- Number
- 1
- Start Page
- 63
- End Page
- 70
- URI
- https://scholarworks.bwise.kr/gachon/handle/2020.sw.gachon/91566
- DOI
- 10.4283/JMAG.2024.29.1.63
- ISSN
- 1226-1750
2233-6656
- Abstract
- This study aimed to confirm the characteristics of deep-learning image reconstruction (DLIR) intensity in abdominopelvic computed tomography (CT) using noise level and blind quality evaluation parameters. The study was conducted using phantoms and patients, and CT images were obtained while adjusting the intensity of DLIR to low (DLIR-L), middle (DLIR-M), and high (DLIR-H). To quantitatively evaluate image quality, the coefficient of variation (COV) and contrast-to-noise ratio (CNR), as well as natural image quality evaluation (NIQE) and blind/referenceless image spatial quality evaluator (BRISQUE), were used. In both the noise level and blind quality evaluation results, a higher strength of DLIR resulted in better results derived from the phantom and patient studies. In particular, the results of the phantom study confirmed that NIQE and BRISQUE of CT images acquired using DLIR-H were improved by approximately 5 % and 1 %, respectively, compared to the corresponding the application of DLIR-L. Moreover, when high-strength deep-learning was applied to a real patient's CT image reconstruction method, the NIQE and BRISQUE results improved by approximately 6 % and 4 %, respectively, compared with their respective medium levels. In conclusion, we quantitatively analyzed the image quality according to the intensity of the recently developed deep-learning-based CT image reconstruction method.
- Files in This Item
- There are no files associated with this item.
- Appears in
Collections - ETC > 1. Journal Articles
![qrcode](https://api.qrserver.com/v1/create-qr-code/?size=55x55&data=https://scholarworks.bwise.kr/gachon/handle/2020.sw.gachon/91566)
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.