Premium
Neurotransmission SPECT and MR registration combining mutual and gradient information
Author(s) -
Bullich S.,
Ros D.,
Pavía J.,
Cot A.,
López N.,
Catafau A. M.
Publication year - 2009
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.3232002
Subject(s) - mutual information , image registration , artificial intelligence , computer science , single photon emission computed tomography , similarity (geometry) , computer vision , image resolution , magnetic resonance imaging , pattern recognition (psychology) , spect imaging , nuclear medicine , image (mathematics) , medicine , radiology
Purpose: Image registration is important in functional image analysis. In neurotransmission single photon emission tomography (nSPECT), specific uptake sites can be accurately localized by superimposing the SPECT study onto a high‐resolution structural image such as a magnetic resonance (MR) of the subject. Mutual‐information (MI)‐based algorithms are usually employed for this purpose. Nevertheless, nSPECT/MR registration using MI is often limited by the low count rates present in nSPECT. Several works have proposed extensions of the MI measures to include gradient information (GI) from the images but their performance has not been evaluated in SPECT studies. Methods: In this work, the accuracy of the MI including gradient information (MIG) was compared with the standard MI using data from healthy volunteers and data simulating a specific uptake reduction using three different radioligands: I123 ‐ IBZM ,I123 ‐ ADAM ,I123 ‐ R 91150 . Results: The results showed that MIG‐based registration yielded better accuracy than MI. The MIG‐based similarity measures were less sensitive to sparse sampling and diminished computational time without a substantial decrease in registration accuracy. Conclusions: Accuracy of nSPECT/MR registration is improved when gradient information is included in the MI‐based algorithm, which makes MIG‐based registration potentially useful for clinical applications.