Published in:
01-04-2005 | Original Article
Diagnosis of Alzheimer’s disease using brain perfusion SPECT and MR imaging: which modality achieves better diagnostic accuracy?
Authors:
Takao Kubota, Yo Ushijima, Kei Yamada, Chio Okuyama, Osamu Kizu, Tsunehiko Nishimura
Published in:
European Journal of Nuclear Medicine and Molecular Imaging
|
Issue 4/2005
Login to get access
Abstract
Purpose
The purpose of this study was to compare the accuracy of MR imaging and brain perfusion single-photon emission tomography (SPECT) in diagnosing Alzheimer’s disease (AD).
Methods
The transaxial section display of brain perfusion SPECT, three-dimensional stereotactic surface projection (3D-SSP) SPECT image sets, thin-section MR imaging of the hippocampus and perfusion MR imaging were evaluated in 66 subjects comprising 35 AD patients and 31 subjects without AD. SPECT and MR imaging were visually interpreted by two experts and two novices, and the diagnostic ability of each modality was evaluated by receiver operating characteristic (ROC) analysis.
Results
In the experts’ interpretations, there was no significant difference in the area under the ROC curve ( A
z
) between 3D-SSP and thin-section MR imaging, whereas the A
z
of transaxial SPECT display was significantly lower than that of 3D-SSP (3D-SSP: 0.97, thin-section MR imaging: 0.96, transaxial SPECT: 0.91), and the A
z
of perfusion MR imaging was lowest (0.63). The sensitivity and specificity of each modality were, respectively, 80.0% and 96.8% for 3D-SSP, 77.1% and 96.8% for thin-section MR imaging, 60.0% and 93.5% for transaxial SPECT display and 34.3% and 100% for perfusion MR imaging. In the novices’ interpretations, the A
z
, sensitivity and specificity of 3D-SSP were superior to those of thin-section MR imaging.
Conclusion
Thin-section hippocampal MR imaging and 3D-SSP image sets had potentially equivalent value for the diagnosis of AD, and they were superior to transaxial SPECT display and perfusion MR imaging. For avoidance of the effect of interpreters’ experience on image evaluation, 3D-SSP appears to be optimal.