Purpose: To evaluate a fully automatic computer-assisted diagnostic system for mild dementia with Lewy bodies (DLB), permitting distinction from mild Alzheimer's disease (AD).
Methods: Using 18F-fluorodeoxyglucose and positron emission tomography (FDG-PET), glucose metabolic images were obtained from mild DLB and mild AD patients. Two groups consisting of 16 mild DLB patients and 21 mild AD patients were recruited for diagnostic evaluation between mild DLB and mild AD. The mean age+/-SD of the mild DLB group and the mild AD group was 74.3+/-4.9 and 71.7+/-2.1 years, respectively, and the mean scores of the MMSE for the mild DLB and the mild AD group were 21.7+/-1.9 and 23.1+/-2.1, respectively. A receiver operating characteristic (ROC) analysis was performed to compare the diagnostic performance, in terms of discrimination between DLB and AD, of conventional axial FDG-PET images inspected visually by experts and beginners with that of our fully automatic diagnosis system using the statistical brain mapping method and Z scores obtained with the DLB template.
Results: The diagnostic performance of the automatic system was comparable to that of visual inspection by experts. The area under the ROC curve for the automatic diagnosis system was 0.77. The mean area under the ROC curve for visual inspection by experts and beginners was 0.76 and 0.65, respectively.
Conclusion: The fully automatic differential diagnosis system for distinction between mild DLB and AD showed a similar diagnostic accuracy to visual inspection by experts. It would be a useful diagnostic tool to distinguish mild DLB from mild AD in clinical practice.