| Abstract No.: | B-F2175 |
| Country: | Canada |
| | |
| Title: | A SPECT CEREBRAL BLOOD FLOW STUDIES RECONSTRUCTION TECHNIQUE CONSTRAINED BY ANATOMICAL INFORMATION FROM MRI SIGNIFICANTLY IMPROVES QUANTIFICATION ACCURACY |
| | |
| Authors/Affiliations: | 1 Jean-Paul Soucy*; 2 Said Benameur; 2 Jean Meunier; 2 Max Mignotte;
1 CHUM/MNI, Montreal, QC, Canada; 2 DIRO, Université de Montréal, QC, Canada
|
| | |
| Content: | Cerebral blood flow Single Photon Emission Computed Tomography (SPECT) studies using “fixed-distribution” agents such as 99mTc-HMPAO or 99mTc-ECD represent a widely available, non-invasive and inexpensive approach to assess on a regional basis cerebral “activity” distribution, defined as (mostly) glutamatergic synaptic exchanges intensity, making them useful in a variety of clinical and research settings. Unfortunately, these 3D data sets are severely degraded by the high noise level they exhibit, which is a consequence of physical factors such as the Poisson nature of nuclear decay, attenuation and scattering of radioactivity within the subject being imaged or within different components of the imaging system, as well as electronic noise of that system itself. This leads to substantial errors in measurements of regional brain blood flow distribution, as well as in absolute quantification of that parameter based on SPECT, and therefore in our estimations of brain activity. We already have shown the potential usefulness of image restoration based on the NAS-RIF algorithm in that context (SfN 2001). Here, we propose a new approach based on those initial results, which constrains the reconstruction process with high resolution anatomical information extracted from an MRI study obtained in the same patient. In order to do this, we incorporate a regularization term which significantly stabilizes the inverse solution to be recovered. That anatomy-based regularization term uses the result of an unsupervised Markovian segmentation obtained after a preliminary registration step based on the maximisation of mutual information from the MRI and SPECT data volumes of the patient. So far, this method has been successfully tested on 30 pairs of brain MRI and SPECT acquisitions from different subjects and on Hoffman and Jaszczak SPECT phantoms acquisitions. When compared with a standard restoration approach using a Metz filter, our method is clearly superior in terms of signal-to-noise ratio: the average ISNR (Improvement in Signal-to-Noise Ratio) of our technique is measured at 0.7 dB, with the Metz filter generating a 0.4 dB SNR, as assessed with the known distribution of the Hoffman brain phantom. We also used another specific evaluation criterion including estimations of the average contrast of the image and the image mottle (a parameter which estimates noise) in different regions of the phantom: our method enhances contrast of the image with very little increase in mottle, i.e., without significant noise amplification, avoiding the creation of artefactual features. Finally, a very important property of our algorithm is that no information about the Point Spread Function of the SPECT imaging process is used, while other filtering approaches require such information. Therefore, this approach can significantly improve SPECT studies of cerebral blood flow by allowing for a quantitatively much better depiction of the distribution of blood flow. |
| | |
| Back |
|
