Reducing motion artifacts in 4D CT image using principal component analysis combined with linear polynomial fitting model
Download as PDF
DOI: 10.23977/AICT2020020
Author(s)
Guangpu Shao, Xiaokun Hu, Jimin Yang and Juan Yang
Corresponding Author
ABSTRACT
We propose a mathematical approach to reduce irregular artifacts in 4d ct images by fitting the generated displacement vector field (dvfs) from deformable image registration(dir) in three temporal and one spatial dimensions through a linear polynomial fitting model, and then use principal component analysis (pca) to decompose the fitted dvfs into a linear combination of the main motion basis to represent conventional respiratory motion. The “Synthetic” ct image of the selected phase is generated by deforming the reconstructed dvfs with the reference ct 00 image. Preliminary results show that this method has the potential to extract regular breathing movements from patients' 4d ct images, and can recover tumor organs and tissues caused by irregular breathing movements during 4d ct image acquisition. Calculate the correlation coefficients(cc) and mean of five patients. Of all the patients ,at superior-inferior(si), anterior-posterior(ap) and medial-latera(ml) mean values of 0.86±0.02, 0.86±0.02 and 0.86±0.02, respectively.
KEYWORDS
4d ct;pca; linear polynomial fitting; dvfs