IR utilizes the collected data and breaks it down into pixels and then treat each pixel value as an unknown. Using an algorithm values can be assigned to each pixel, permitting the emission and detection data to be accurately modeled.
Noise acquired during the SPECT acquisition, in the absence of attenuation correction, is constant across the reconstructed field. This eliminates streaking artifact in the image and provides a visually superior image to filtered back projection (FB) processing.
The above image of a bone whole body spect, processed with iterative reconstruction shows a large photopenic area where the femurs should be. Using IR the pixel values of a low count area may be treated as background or noise in comparison to the adjoining pixels...
Filtered backprojection, takes the acquired data and fills in the spaces between the projections (azimuths) to create a completed 360-degree image. Although this completes the image it does create blurring, to which filters can be applied for an improved appearance. Unlike IR processing, FB does contain noise and background activity..
Recprocessing the acquired data with filtered back projection shows a complete if not more favorable image. Note the noise or "rays" of activity surrounding the scan, especially around the bladder activity...
References:
1. Image reconstruction - a tutorial, G.L. Zeng, Computerized Medical Imaging and Graphics 2001
2. An introduction to iterative reconstruction, B Hutton
3. Filtering in SPECT image reconstruction, M. Lyra and A. Ploussi, International Journal of Biomedical Imaging, April 2011
