Medical Imaging

Lab 2:
Computed
Tomography (CT)
Le Nhat Tan MEng.
Department of Biomedical Engineering
Faculty of Applied Science
HCMUT - VNUHCM
[email protected]
Content

1. Review
2. Tomography Reconstruction Algorithm
1. S1: Filtered Back Projection (FBP)
2. S2: Iterative Reconstruction (SIRT)
3. S3: Iterative Reconstruction (CGLS)

2
1. Review
1.1. What is CT? (1)
Computed Tomography
- Image Reconstruction: images are created • Imaging techniques that generate detailed
through a mathematical process. cross-sectional images (or slices) of a specific
part of the body.
- The "computation" involves reconstructing • The term emphasizes the creation of images of
images by using information from multiple X- thin, cross-sectional "slices" of the body,
ray projections correspond to multiple angle. making it possible to examine structures in
great detail
 1. Review
1.2. Principle (1) – X-ray tube in CT

Pencil beam fan-beam cone-beam

Fan-shaped (Captures a single slice per rotation) and cone-

shape (Captures a volume per rotation) X-ray beam [1]

[1] Adam Thompson et al, Measurement Science and Technology 27(7), 2001
 1. Review
1.2. Principle (2) - Detectors

Single-Slice Detectors (SSD) and Multi-

Slice Detectors (MSD) [2]
[2] Armin Schneider, Hubertus Feussner, in Biomedical Engineering in
Gastrointestinal Surgery, 2017 Spectral photon-counting CT scanners [3]
[3] David P. Cormode et al, Scientific Reports 7, 4784, 2017.
 1. Review
1.2. Principle (3) – Gantry Rotation

The Slip Ring

uses gliding contacts (like
brusher) to allow
communication
and power transfer
between the stationary
and rotating frames
without the use of wires.

There are also numerous small tracks of the slipring that are
used to conduct low-power (digital) signal data off of the
rotating gantry to the stationary frame.

[4] Radiology Key, Computed Tomography, Access at: https://radiologykey.com/computed-tomography-15/

 2. Tomography Reconstruction Algorithm
2.1. Projection and Back Projection (1)
• A sinogram is a 2D representation of
the raw projection data gathered
during a scan.

• Each projection angle produces its

own sinogram.

[5] imec - Vision Lab, Filtered Backprojection (FBP)

[6] Seibert, J. A., Pediatric Radiology, 44(S3), 431–439, 2014.
 2. Tomography Reconstruction Algorithm
2.1. Projection and Back Projection (1)

Object

[5] imec - Vision Lab, Filtered Backprojection (FBP)

 2. Tomography Reconstruction Algorithm
2.1. Projection and Back Projection (2)

• Back-projection is the
process of reconstructing
an image by combining all
these sinograms by
adding the projections
together.

-> However, the result of

back-projection is often a
blurry image.

[6] Dejian Qiu and Euclid Seeram, Open Journal of Radiology 1(2):42-54, 2016
2. Tomography Reconstruction Algorithm
2.1. Projection and Back Projection: Practice

1. Load Images/Create Sample Images

2. Stimulate the Projection (by Radon Algorithm)

3. Perform the Back Project

4. Visualization
 2. Tomography Reconstruction Algorithm
2.2. Filtered Back Projection

[5] imec - Vision Lab, Filtered Backprojection (FBP)

2. Tomography Reconstruction Algorithm
2.2. Filtered Back Projection: Filters
The process of filtering in FBP
follows these steps:
• Convert the projection data
into the Fourier domain
using the Fast Fourier
Transform (FFT).
• Apply the chosen filter
function in the frequency
domain.
• Convert the filtered projection
back to the spatial domain
using the Inverse FFT.
• Perform back projection to
reconstruct the final image.
 2. Tomography Reconstruction Algorithm
2.2. Filtered Back Projection: Fourier domain
The process of filtering in FBP follows
these steps:
• Convert the projection data into the
Fourier domain using the Fast
Fourier Transform (FFT).
• Apply the chosen filter function in
the frequency domain.
• Convert the filtered projection back
to the spatial domain using the
Inverse FFT.
Low: smooth surface • Perform back projection to
High: details and the sharp edges
reconstruct the final image.
[5] imec - Vision Lab, Filtered Backprojection (FBP)
2. Tomography Reconstruction Algorithm
2.2. Filtered Back Projection: Practice
1. Load Images/Create Sample Images
2. Stimulate the Projection (by Radon Algorithm)
3. Perform the FFT, apply filters
4. Back Projection
5. Visualization
 2. Tomography Reconstruction Algorithm
2.4. Iterative Reconstruction

Iterative Reconstruction refines the reconstructed image over multiple iterations,

leading to improved image quality, better noise handling, and lower radiation dose
requirements.
[6] Seibert, J. A., Pediatric Radiology, 44(S3), 431–439, 2014.
 2. Tomography Reconstruction Algorithm
2.4. Iterative Reconstruction: SIRT
1. Initial Estimate: Start with an initial guess
of the image (usually a uniform image or FBP-
reconstructed image).
2. Forward Projection: Compute the
projection of the estimated image.
3. Compare with Measured Data: Calculate
the difference (error) between the estimated
and the actual measured projections.
4. Update the Image Estimate: Modify the
image based on the error using a
mathematical model.
5. Repeat Until Convergence: Steps 2–4 are
repeated until the solution converges or
meets a stopping criterion (e.g., error
Statistical Iterative Reconstruction (SIRT) becomes small).
[7] Shabab Bazrafkan et al, Preprint Arvix 10.48550/arXiv.1910.06565
 2. Tomography Reconstruction Algorithm
2.4. Iterative Reconstruction: CGLS
CGLS (Conjugate Gradient Least Squares)
is a type of model-based iterative
reconstruction that solves the inverse
problem using an optimization approach.

• Its goal is to find an image that minimizes

the difference (squared difference)
between the measured and simulated
projections , incorporating a model of the
imaging system.

• Conjugate Gradient: converges to the

Model-based Iterative Reconstruction least-squares solution by moving along
CGLS (Conjugate Gradient Least Squares)
carefully chosen directions
[8] Lu Liu Honours, Journal of Medical Imaging and Radiation Sciences Volume 45, Issue 2, June 2014, Pages 131-136
 2. Tomography Reconstruction Algorithm
2.3. Iterative Reconstruction

[6] Seibert, J. A., Pediatric Radiology, 44(S3), 431–439, 2014.

[8] Davide Micieli et al, SoftwareX 9 :260-264, 2019
2. Tomography Reconstruction Algorithm
2.3. Iterative Reconstruction: Practice
1. Load Images/Create Sample Images
2. Stimulate the Projection (by Radon Algorithm)
3. Perform the Iterative Reconstruction (SIRT and
CGLS)
4. Visualization
3. Report
• Application