Image acquisition Flashcards
What is frame mode imaging ?
Image parameters are set before imaging, usually stored as protocols, i.e. mat size, mag, mat depth, e window and termination condition
What is list mode imaging?
Data is stored as x,y,z and t Can rebin/frame as desired (flexible) Inc post-processing time No check Im during imaging Option not always available
What is static imaging?
Images are displayed as a matrix
Little or no post-pro
Take diff angle views to get more info
Why is whole body imaging used?
Easier to interpret than multiple statics
How are whole body images created?
1)Step and shoot:
Series of static images
Digitally stitched together
2) Single sweep
Scan speed pre-selected
Counts stored in pos mat based on cam position at the time of recording
What is dynamic imaging?
A series of sequential static frames
Used to image rapidly changing distribution within the patient
ex. renography
What is cardiac gated imaging?
Several frames acquired covering the cardiac cycle
Acquired over many cycles
Can bin counts based on phases
What is SPECT?
3D GC projection imaging
What are the issues with 2D imaging?
2D view of a 3D object
There is no depth information
Overlying structures are superimposed leading to a loss of contrast compared to 3D
How are planar images motion corrected? (4 steps)
1 Produce line images from a dynamic sequence (compress along each row/clm)
2 Stack the line images and est x/y motion from the deviations in pos of max I
3 Translate each frame according to the disp co-ords
4 Sum the translated frames
How are SPECT images motion corrected?
Review sinograms and linograms for discontinuities
The shift req to realign the components is then calculated and the projections are translated
What are the limitations of motion correction?
Reduces spatial res as req interpolation
Out of plane motion cannot be corrected
Distribution changes can confound the algorithms
How are colour scales used?
Help simplify diagnosis by maximising visible contrast
Describe Filtered Back Projection
Back Project each planar image onto three dimensional image matrix
More views – better reconstruction
1/r blurring, even with infinite number of views
So must apply a ramped filter to remove the blur
How does using a ramp filter effect projection images?
Gives increases weighting to higher spatial frequencies
But, amplifies noise
How can ramp filters be improved?
Multiply with another function to rolls off the ramp effect at higher frequencies
For example, mult by a gaussian
What are the limitations of back projection?
Requires filtering to correct
Can introduce noise and streaking artefacts
Filtering can reduce issues but reduces resolution
What are the 2 types of attenuation correction?
Chang and Sorenson
What is Sorenson correction?
Assumes known uniform attenuation, with object size estimate from an orthogonal projection.
Modifies projections before reconstruction
What is Chang correction?
1) Assumes known uniform attenuation over the object and requires object boundary is defined a priori.
2) Reconstructs the uncorrected image
3) Calculate correction matrix and applies it as a scaling factor
What is second order chang?
1) Uses first order result to create new, corrected projections
2) Subtracts the corrected projections from acquired data to give an ‘error’ projection set,
3) Then the ‘error’ set is reconstructed, subjected to first order correction and added to corrected image
What are the limitations of attenuation correction?
Inaccurate corrections can cause artefacts
Usually designed for use without scatter correction u values should be broad beam
Assumes uniform atten, good for soft tissues/bone poor for lung/thorax
What is the minimum number of projections required to avoid streaking?
Need sampling ∆𝜃≤𝑑/(1 ⁄ 2 𝐷)
For 360 ∆𝜃=2𝜋/𝑁 so 𝑁≥𝐷/𝑑
For 180 N/2 will do
min is approx same no views as matrix size
What is iterative reconstruction?
Iteration is process of successively better “guesses”
Creates an image by refining the expected projections in comparison to those recorded MLEM
Minimum N of iterative projections required?
Within a circle of diameter D there will be
𝜋 ((𝐷⁄𝑑))⁄4 pixels D=diam d=pixel sep
Ignoring atten and res loss N/2 proj provide new info (in 360)
Therfore:
𝑁/2 𝐷/𝑑≥𝜋/4 (𝐷/𝑑)^2 𝑜𝑟 𝑁≥𝜋/2 𝐷/𝑑
less than for FBP
What are the main Iterative algorithms?
Conjugate Gradient
Maximum Likelihood Expectation Maximisation (MLEM) (and its offshot OSEM)
Maximum A Posteriori Expectation Maximisation (MAPEM)
What is MLEM?
Each iteration of the algorithm has two steps:
1) In the expectation step the formula giving the likelihood of any reconstructed image given the measured data is formed
2) In the maximisation step the image that is most likely given the measured data is found
What is OSEM ?
Ordered Subsets Expectation Maximisation
OSEM accelerates convergence by updating the image estimate following a comparison of only a sub-set of the projection data to the current estimated projections
Reduces comp demand, inc convergence by factor of the no of sub-sets used
Must carefully select subsets to obtain best convergence
Possible refinements to atten corr?
- Integrate a system specific correction into the initial projection
- Can use CT data to provide an atten corr map
How is scatter correction used?
Not widely available
Example is monte carlo
Atten corr must account for scatter corr
What are the advantages of SPECT/CT?
CT can provide an accurate atten corr map even with low dose/qual CT
Simplifies location of lesions by providing improve co-registration of the modalities - full dose CT not required
Faster diag/staging - one stop imaging
CT atten corr allow for accurate tracer uptake quantisation
Issues with image fusion ?
SPECT-CT images are acquired sequentially not concurrently - swaps gantries
But patient motion betw ims can cause alignment issues esp. breathing issues - must check reg with phantom ims
