Application of fast non-local denoising approach in digital radiography using lung nodule phantom for radiation dose reduction

Document Type : Original Paper

Authors

1 Department of Diagnostic Radiology, Severance Hospital, Seoul, Republic of Korea

2 Department of Bio-Convergence Engineering, Korea University, Seoul, Republic of Korea

3 Department of Radiological Science, Gachon University, Incheon, Republic of Korea

Abstract

Introduction : Chest X-ray imaging has become the most commonly used, as it is the primary method for lung cancer screening during medical check-ups. The radiation dose should be minimized to ensure that the patients are not overexposed to radiation. However, radiation dose reduction results in increased noise in the chest X-ray image. Thus, the purpose of this study was to evaluate utility of fast non-local means (FNLM) filters to reduce radiation dose while maintaining sufficient image quality.

Materials and Methods : This study evaluates three filters (median, Wiener, and total variation) and a newly proposed filter (fast non-local means (FNLM)), which reduce image noise. A realistic anthropomorphic phantom is used to compare image acquired depended on positions such as anterior-posterior, lateral, and posterior-anterior, using self-produced 3D printed lung nodule phantom. To evaluate image quality, we used the normalized noise power spectrum (NNPS), contrast to noise ratio (CNR), and coefficient of variation (COV) evaluation parameters.

Results : The NNPS and COV was lowest and the CNR was highest with FNLM images. FNLM filter outperforms other compared filters in terms of noise reduction.

Conclusion : Therefore, use of an FNLM filter is recommended, because it reduces the radiation dose to a patient and thus minimizes the risk of cancer, while maintaining diagnostic quality.

Keywords

Main Subjects



Articles in Press, Accepted Manuscript
Available Online from 25 January 2022
  • Receive Date: 28 August 2021
  • Revise Date: 19 January 2022
  • Accept Date: 25 January 2022
  • First Publish Date: 25 January 2022