CT Metal Artifact Reduction Method Based on Improved Image Segmentation and Sinogram In-Painting

Table 1

Parameter setting for Bal’s algorithm and our proposed method. For our proposed correction, MCN_𝑎 and MCN_𝑚 depict the maximum class number (MCN) in the MIMS segmentation of artifacts and metal components, respectively.

		ClinicalImage1	ClinicalImage2	ClinicalImage3	PhantomImage
Bal algorithm in [7]	Prefilterig step	𝜐=2, the scaling factor 𝜎0 is set to 10,
		the relationship between width and length of Gaussian filter is set to 2 (𝜎⟂=2𝜎‖).
	Segmentation step	K-means segmentation using 5 classes with CT values: −950 (air),
		0 (soft tissue), 200 (normal tissue), 750 (bone), 5000 (metal).
	Inpainting step	Linear interpolation using 5 points in each symmetric side
Our proposed method	Pre-filterig step	41×41𝑁,11×11𝑠,ℎ=500	41×41𝑁,11×11𝑠,ℎ=450	41×41𝑁,11×11𝑠,ℎ=450	41×41𝑁,11×11𝑠,ℎ=350
	Segmentation step	MCN𝑎=16,MCN𝑚=22,; 𝜀=600	MCN𝑎=12,MCN𝑚=18,; 𝜀=600	MCN𝑎=12,MCN𝑚=18,; 𝜀=600	MCN𝑎=5,MCN𝑚=7,; 𝜀=600
	In-painting step	11×11P^𝑝,3×3P^𝑖𝑞,𝜆=0.65	11×11P^𝑝,3×3P^𝑖𝑞,𝜆=0.3	11×11P^𝑝,3×3P^𝑖𝑞,𝜆=0.3	11×11P^𝑝,3×3P^𝑖𝑞,𝜆=0.8