Iterative Reconstruction Flashcards
what is a filtered back projection
it is a closed form formula. One step solution, computationally efficient. Exact reconstruction if applied to ideal projection data
- point source
- point detector
- x-ray path as a thin line
- monochromatic spectrum
- complete data set
what is the size of xray source and detector in FBP
it is an finite size of source and detector
what is the spectrum for FBP
it is polychromatic spectrum, beam hardening
what is the noise for FBP
it is quantam noise, and electronic noise
what is the projection data like
it is an incomplete projection data
what is iterative reconstruction- what is the initial guess
it can be anything and it is most commonly a blank image or FBP images
what is the simulated projection in an iterative reconstruction
add attenuation (CT number) together along x-ray path for reconstructed image is correct, measured projection data should be the same as those calculated from reconstructed images
draw out how iterative reconstruction usually goes
slide 3 of the iterative reconstruction power point
what are the advantages of iterative reconstruction
accurate system modeling: geometry, spectrum
appropriate statistcal noise model
may improve statistical resolution and reduce image artifacts
what are the limitations of iterative reconstruction
computationally intensive, long reconstruction time, different image appearance
what is iterative reconstruction used for most of the time
it was used in PET, SPECT, IR in the early 70s, limited applications, reintroduced CT, so some people use it now in CT for IR
is iterative reconstruction used today and if so, how long does it take
it is used today, but reconstruction time is seconds to thirty minutes
do you use raw data or images for iterative reconstruction
it is better to se a picture
draw an iterative reconstruction for te approach
page 6 of the power point
what is algebraic reconstruction
images with n pixels have N degrees of freedom. A specific image is a single point in N dimension space, and single ray represents a hyperplane in this space. Image reconstruction is to find the intersectionpoint of all hyperplanes
what are the steps of iterative reconstruction
initialization forward projection compare with measured projection backproject the difference repeat steps until the criteris is met
when do you stop the criteria
when a fixed number of iterations is met, when the difference between the images in two iterations is smaller than a predetermined threshold
what is a statistical reconstruction
consider measured projections (sinogram) as random variables. The statistics of these random variables are related to the imaging object (estimated images). An objective function is defined to relate measured projections to the estrimated images. IMage reconstruction is performed by iteratively maximized (minimized) the objective function. Stabalizing function in the form of a regularization prior may further reduce artfacts and noise
what is are some of the models for statistical iterative reconstruction
Poisson noise model and maximum a posteriori
what is data fidelity
it is a match measured and simulated projection data. You assign different weights to the different projection data:
- lower weight to higher noise projection data
- higher weight to lower noise projection data
what is the regularization function
without the redularization term, the image estrimates are excessively noisy and unstable. Regularization enforces smoothnessin the reconstructed images by encouraging neighboring pixels to have similar values. Based on certain prior knowledge of the image. Tradeoff between noise and resolution
what are the advantages of iterative reconstruction
accurate geometry model spectrum model appropriate statistical noise model object constraints (non-negativity) artifact reduction (can reduce metal artifacts) flexible for scanning configurations potential to reconstruct images from sparse or incomplete data
what is the spectrum model
it is for beam hardening correction by modeling beam spectrum into iterative reconstruction
what is the difference between FBP and iterative reconstruction in a statistical model
FBP treats all data equally
iterative reconstruction- weight is measured data accordingly, assign lower weight for noisier data, potential reduction and dose reduction
what is better for metal artifact reduction
it is best with iterative reconstruction
is FBP or iterative based on geometry
FBP has an algorithm for each, but a new one has to be employed for each different geometry. Iterative reconstruction treats it all the same so its flexible for non-standard geometry
does noise reduction improve detection
multiple studies have shown noise reduction does improve image quality and diagnositc confidence, however low dose images withou tnoise reduction show the same CT findings
what is observer performance
multiple studies evaluate image quality of lower dose images with nise reduction, not observer performance, and comaprison with full dose images to determine what you are missing, and comparison with low dose images, to determine if you need to noise reduction to change diagnosis
what is the lowest acceptable dose
Need an observer for it, and it is better be safe leaving a safety margin. Higher contrast tasks permit greater dose reduction, while low contrast tasks permit less.
what is the ideal observer
Bayesian
what is the ideal linear observer
Hotelling observer
what is model observers that predict human observers that predict human observer perfromance
anthropomorphic
