 |
 |
If you are not already referring to it, then I would suggest taking a look at: Richardson et. al. (1997). "Cortical grey matter and benzodiazepine receptors in malformations of cortical development. A voxel-based comparison of structural and functional imaging data. Brain. 120:1961-1973 | We are working on atrophy correction based on T1 values of individual | voxels This is outside of SPM. We wonder if atrophy correction could be | done inside SPM by entering PET/SPECT and T1 MR images without the | correction we are working on. The resolution of T1 MRI will be adjusted | to the resolution of PET/SPECT images based on point-spread function for | individual voxels. Therefore, spatial resolution of PET/SPECT and the | resolution-adjusted T1 MRI is the same for individual voxels. | | Let's assume that we compare patients and healthy controls. Patients may | show decreased uptake of radiotracer but it may be due to the decrease in | gray matter volume. We want to separate the source of the decreased uptake. | It may be due to the decreased uptake per gray matter volume, decreased | gray matter volume without change in the uptake per gray matter volume, or | both. | | Enter four sets of images, 1. patients' PET/SPECT, 2. controls' PET/SPECT, | 3. patients' resolution-adjusted MRI, and 4, controls' resolution-adjusted | MRI. | How can we interpret the results by the contrast | -1+1-(-1+1)? | Will this contrast pick up the regions where the decrease in tracer uptake | is significantly greater than the decrease in gray matter volume? It should. | However, the noise of the images is of course different. PET/SPECT images | are much noisier. We wonder if images with different noise levels could be | entered in the same analysis. Or do we need to add noise to the | resolution-adjusted T1 MRI? After smoothing, the noise should not be too much of a problem as most of the residual variance should arise because the spatial normalization is not exact. These errors should be the same for both the PET/SPECT and the smoothed grey matter images. You will probably need to be more concerned about the global intensity normalization, since what you really want to test is pat_SPECT/pat_GM - con_SPECT/con_GM. This is inherantly unstable, so an interraction between the SPECT intensity and the smoothed GM is used instead. The optimum global normalization would make the SPECT and GM intensities as similar as possible. Sub-optimum globals may bias your results. | Another concern is that spatial normalization may partially correct gray | matter atrophy. In the nonlinear algorithm, thin gray matter could become | thicker? Do we need to partially disable nonlinear algorithm in spatial | normalization? The grey matter could become thicker because of the nonlinear normalization, but it should become equally thick in both the SPECT and the smoothed GM images, so should not introduce additional interraction effects. Note that even an affine normalization can make the cortex thicker, simply by globally stretching a brain. When doing any group comparisons, there will inevitably be some side effects from the spatial normalization. For example, if one group has smaller brains than the other, then the spatial normalization will enlarge the smaller brains, thus increasing the cortical thickness. This may, or may not be what you want to happen. Good luck, -John %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%