Evaluation of a new expert system for fully automated detection of the Alzheimer's dementia pattern in FDG PET

Objective Fluorodeoxyglucose (FDG) positron emission tomography (PET) is increasingly used to support a diagnosis of Alzheimer's disease. The aim of the present study was to evaluate a new expert system (PALZ) for the fully automated analysis of FDG PET images for diagnosis of the disease. Methods The PALZ tool is based on the detection of the typical disease pattern in FDG PET images. Its potential for this task was evaluated in 22 consecutive patients with suspected Alzheimer's disease who had been graded as positive for the pattern by an experienced reader (visual analysis supported by statistical parametric mapping (SPM)), and in 18 controls. Dependence on scanner performance was assessed by variation of the spatial resolution of the PET images. Results All the Alzheimer's disease subjects were classified as pattern-positive by the PALZ tool. Fifteen controls were classified as normal. Sensitivity and specificity for differentiation of the patients with suspected Alzheimer's disease from the controls were 100% and 83%, respectively. The false positive finding in three controls most likely was caused by differences in attenuation correction between the normal data base of the PALZ tool (cold transmission scan) and the local data sets (hot transmission scan). There was only mild dependence on spatial resolution. Conclusions The results of the present study suggest that the PALZ tool provides similar performance for the detection of the typical Alzheimer's disease pattern in FDG PET images as an experienced reader supported by SPM. The PALZ tool is fully automated, easy to use, and insensitive to the spatial resolution of the PET scanner used. Therefore, it has the potential for widespread clinical use.