OBJECTIVES: Alzheimer's disease (AD) is diagnosed by either inspection of the brain perfusion SPECT, or three-dimensional stereotactic surface display (3D-SSP). The purpose was to compare diagnostic performances of these methods. METHODS: Sixteen nuclear medicine physicians independently interpreted 99mTc-ECD SPECT in one session and SPECT with 3D-SSP in another session without clinical information for 50 studies of AD patients and 40 studies of healthy volunteers. Probabilities of AD were reported according to a subjective scale from 0% (normal) to 100% (definite AD). Receiver operating characteristics curves were generated to calculate areas under the ROC curves (Az's) for the inspection as well as for an automated diagnosis based on a mean Z value in the bilateral posterior cingulate gyri in a 3D-SSP template. RESULTS: Mean Az for visual interpretation of SPECT alone (0.679 +/- 0.058) was significantly smaller than that for visual interpretation of both SPECT and 3D-SSP (0.778 +/- 0.060). Az for the automated diagnosis (0.883 +/- 0.037) was significantly greater than that for both modes of visual interpretation. CONCLUSIONS: 3D-SSP enhanced performance of the nuclear medicine physicians inspecting SPECT. Performance of the automated diagnosis exceeded that of the physicians inspecting SPECT with and without 3D-SSP.
OBJECTIVES:Alzheimer's disease (AD) is diagnosed by either inspection of the brain perfusion SPECT, or three-dimensional stereotactic surface display (3D-SSP). The purpose was to compare diagnostic performances of these methods. METHODS: Sixteen nuclear medicine physicians independently interpreted 99mTc-ECD SPECT in one session and SPECT with 3D-SSP in another session without clinical information for 50 studies of ADpatients and 40 studies of healthy volunteers. Probabilities of AD were reported according to a subjective scale from 0% (normal) to 100% (definite AD). Receiver operating characteristics curves were generated to calculate areas under the ROC curves (Az's) for the inspection as well as for an automated diagnosis based on a mean Z value in the bilateral posterior cingulate gyri in a 3D-SSP template. RESULTS: Mean Az for visual interpretation of SPECT alone (0.679 +/- 0.058) was significantly smaller than that for visual interpretation of both SPECT and 3D-SSP (0.778 +/- 0.060). Az for the automated diagnosis (0.883 +/- 0.037) was significantly greater than that for both modes of visual interpretation. CONCLUSIONS: 3D-SSP enhanced performance of the nuclear medicine physicians inspecting SPECT. Performance of the automated diagnosis exceeded that of the physicians inspecting SPECT with and without 3D-SSP.