Direct comparison of spatially normalized PET and SPECT scans in Alzheimer's disease

Hexamethylpropyleneamine oxime (HMPAO) SPECT and 18F-FDG PET depict similar aspects of perfusion and metabolic abnormalities in Alzheimer's disease (AD), but the correspondence between them is not known in detail. We therefore used statistical parametric mapping to detect and compare abnormal brain areas objectively and quantitatively.

Twenty-six patients with probable AD (mean age +/- SD, 66 +/- 9 y; mean Mini-Mental State Examination score, 22.5 +/- 4.2) and 6 nondemented healthy volunteers (mean age, 63 +/- 11 y) were studied with HMPAO SPECT and 18F-FDG PET. All images underwent the same processing steps, including 12-mm gaussian smoothing, spatial normalization, and z transformation with respect to normal average and SD. Thresholding of z maps was used to detect abnormal voxels.

The overall correlation between PET and SPECT across the entire brain was significant but not close (average r = 0.43). The best correspondence was found in the temporoparietal and posterior cingulate association cortices. There, the number of abnormal voxels for PET correlated strongly with the number for SPECT (r = 0.90 at a z threshold of -2.25), but tracer uptake reductions were significantly more pronounced for PET than for SPECT. Discordant findings were most frequently seen in the temporobasal and orbitofrontal areas (PET low, SPECT high) and in the cerebellum, parahippocampal cortex, and midcingulate cortex (PET high, SPECT low). The correlation between dementia severity and the number of abnormal voxels was closer for PET than for SPECT. Separation of patients from healthy volunteers by counting the number of abnormal voxels was possible over a much wider range of z thresholds with PET than with SPECT.

Correspondence between 18F-FDG PET and HMPAO SPECT is limited to the main finding of temporoparietal and posterior cingulate functional impairment in mild to moderate AD. The distinction between healthy volunteers and patients is less sensitive to threshold selection with PET than with SPECT, and findings in the frontal, temporobasal, and temporomesial cortices and in the cerebellum may differ between the 2 techniques.