What is iterative reconstruction in SPECT?
Iterative reconstruction is a mathematical method used in SPECT to reconstruct tomographic images by repeatedly refining an estimated activity distribution until it best matches the measured projection data.
Unlike filtered back projection, which applies a direct mathematical formula, iterative reconstruction starts with an initial guess of the tracer distribution and repeatedly compares simulated projections from that estimate with the actual measured projections. The image is updated step by step to minimise the difference.
Iterative reconstruction refines an estimated tracer distribution through repeated comparison with measured projections, allowing modelling of attenuation and system response for improved image quality.
Iterative methods allow modelling of attenuation, scatter, and system resolution, producing more accurate and less noisy images than filtered back projection.
Understanding the physics
In SPECT, the measured projections represent the integrated activity along many lines through the body. Reconstruction aims to determine the three-dimensional tracer distribution that could have produced those projections.
Iterative reconstruction begins with an initial estimate of the activity distribution, often uniform. From this estimate, the system mathematically simulates what the projections would look like. These simulated projections are then compared to the actual measured projections.
The difference between the simulated and measured data is used to update the image estimate. This process is repeated multiple times. With each iteration, the reconstructed image becomes a closer approximation of the true distribution.
Common iterative algorithms include:
MLEM (Maximum Likelihood Expectation Maximisation)
OSEM (Ordered Subsets Expectation Maximisation)
OSEM accelerates convergence by dividing projection data into subsets and updating the image more frequently.
A key advantage of iterative reconstruction is that it can incorporate models of:
Photon attenuation within the patient
Compton scatter
Collimator-detector response
Because these physical processes are explicitly modelled, iterative reconstruction improves quantitative accuracy and contrast resolution compared with filtered back projection.
However, increasing the number of iterations can increase image noise. Therefore, reconstruction parameters must balance resolution and noise amplification.
Where this matters clinically
Iterative reconstruction is now widely used in myocardial perfusion imaging and other SPECT applications. It improves lesion detectability, reduces noise, and allows attenuation correction to be incorporated into reconstruction.
Understanding iterative reconstruction helps explain why modern SPECT images appear cleaner and more accurate than those reconstructed using filtered back projection.