Blind deconvolution of human brain SPECT images using a distribution mixture estimation

Thanks to its ability to yield functionally-based information, the SPECT imagery technique has become a great help in the diagnostic of cerebrovascular diseases. Nevertheless, due to the imaging process, SPECT images are blurred and consequently their interpretation by the clinician is often diicult. In order to improve the spatial resolution of these images and then to facilitate their interpretation, we propose herein to implement a deconvolution procedure relying on an accurate distribution mixture parameter estimation procedure. Parameters of this distribution mixture are eeciently exploited in order to prevent overrtting of the noisy data or to determine the support of the object to be deconvolved when this one is needed. In this context, we compare the deconvolution results obtained by the Lucy-Richardson 1 method and by the recent blind deconvolution technique called the NAS-RIF algorithm 2 on real and simulated brain SPECT images. The NAS-RIF performs the best and shows signiicant contrast enhancement with little mottle (noise) ampliication.