Medical Imaging Reconstruction Algorithm Examples (M.I.R.A.G.E)

Introduction

Reconstructing medical images from raw data stored as a commonly used DICOM format

Visualization of Inverse Radon Transform Using Arithmetic Libraries (V.I.R.T.U.A.L)

Keywords: CT Scanners, Inverse Radon Transform, Sinogram, and Image Reconstruction

關進 : CT掃描, 對偶雷登變換, 正弦圖, 圖像重建

キーワード: CTスキャナ, 逆ラドン変換, 画像の再構成, サイノグラム

Introduction

1. CT Inverse Radon Transformation
• Algebraic Reconstruction Technique (ART) -- Iterative Method
• Filtered Back Projection (FBP)
2. CT Reconstruction in 2D Domain
3. CT Reconstruction in 3D Domain

What will be a typical size of CT image slices? Are they always square-shaped? For most clinical applications, resolution of cross-sectional images are set to 512 x 512 pixels and 1024 x 1024 pixels or more for the state-of-the-art CT scanners (research purpose), see reference below. Therefore, it is safe to assume that a CT image slice will always come with a size of 512 x 512 pixels, as such size has been well enthrenced in the radiological community as a standard.

Comparison between ART and FBP

The following table compares each reconstructed image with the original version. Each image has a size of 128 x 128 pixels.

Original Image	ART Reconstructed	FBP Reconstructed

	ART	FBP
Speed	Slow	Fast
Resolution	Poor	Good
Contrast	Good	Poor
Noise	Low	High

CT Reconstruction in 2D Domain

CT Reconstruction in 3D Domain

CT Window Filtering

Coming Soon...

---

References

1. Martin J. Willemink and Peter B. Noel. (2018). "The Evolution of Image Reconstruction for CT - from Filtered Back Projection to Artificial Intelligence". European Radiology download
2. Avinash Kak and Malcolm Slaney. (1988). "Principles of Computerized Tomographic Imaging Chapter 3 and 7". IEEE PRESS. New York download
3. Jerrold T. Bushberg, J. Anthony Seibert, Edwin M. Leidholdt Jr., and John M. Boone. (2011). "The Essential Physics of Medical Imaging, Third Edition". LWW; Third, North American Edition download