A Method to Estimate a Fan-Beam CT NPS Using Two Basis Functions with a Limited Number of Noise Realizations
Presentations
WE-D-TRACK 3-1 (Wednesday, 7/28/2021) 2:00 PM - 3:00 PM [Eastern Time (GMT-4)]
Purpose: Traditionally, noise power spectrum (NPS) is estimated by ensemble averaging of multiple realizations of noise-only images. However, with a limited number of images, the estimation accuracy of NPS becomes lower. Recent works proposed NPS estimation methods using radial 1D-NPS as its basis, but these methods couldn’t reflect the non-stationary noise property of a fan-beam CT image. In this work, we proposed a more accurate NPS estimation method that reflects the non-stationary noise property of a fan-beam CT image.
Methods: Each angular spoke of the 2D-NPS can be expressed as a sum of two basis functions with different cutoff frequencies determined from complimentary projection magnification factors. Therefore, the proposed method uses two basis functions by reflecting view angle dependent variations of cutoff frequencies in each angular spoke of the 2D-NPS. To examine the non-stationary property of NPS in a fan-beam CT image, NPS was estimated at 3 different local regions centered at (-100mm, 0), (0, 0) and (0, -100mm). The weighting factors of the two basis functions were estimated by minimizing the mean squared error (MSE) from the realized NPS image calculated from 10 noise realizations. We compared the estimation accuracy of NPS calculated from 10 noise realizations, previous method, and proposed method. Analytical NPS was used as a reference.
Results: The NPS estimated by the proposed method showed high similarity to analytical NPS, while the NPS estimated by the previous method couldn’t reflect the non-stationary noise property of a fan-beam CT image.
Conclusion: In this work, we proposed a more accurate method using two basis functions to estimate the NPS which reflects the non-stationary noise property of a fan-beam CT image. The NPS estimated by the proposed method showed higher similarity to analytical NPS compared to the NPS estimated by the previous method.
Funding Support, Disclosures, and Conflict of Interest: This work was supported by the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT under Grants 2019R1A2C2084936 and 2020R1A4A1016619.
Handouts
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email