Construction and comparative evaluation of different activity detection methods in brain FDG-PET

Area of Science:

• Neuroimaging
• Nuclear Medicine
• Computational Neuroscience

Background:

• Positron Emission Tomography (PET) with fluorodeoxyglucose (FDG) is crucial for assessing brain glucose metabolism.
• Detecting dementia-related hypometabolism requires robust reference databases for accurate analysis.
• Software like CAD4D, SPM, and NEUROSTAT aid in user-independent dementia detection.

Purpose of the Study:

• To construct and evaluate reference brain FDG-PET databases for three distinct software programs.
• To assess the impact of different stereotactic normalization algorithms on database construction.
• To evaluate the performance of these databases in detecting dementia-related hypometabolism in Alzheimer's disease (AD) patients.

Main Methods:

• Constructed reference brain FDG-PET databases using data from 37 healthy volunteers.
• Compared databases generated by CAD4D, SPM, and NEUROSTAT to evaluate normalization algorithm effects.
• Analyzed FDG-PET scans from 20 AD patients and 10 non-AD patients to assess hypometabolism detection.

Main Results:

• Observed differences in reference databases due to distinct normalization methods, leading to altered spatial FDG patterns.
• All software packages identified similar hypometabolism patterns in AD patients.
• CAD4D and NEUROSTAT reported larger z-scores than SPM; concordance improved with SPM's normalized images and independent z-score calculation.

Conclusions:

• Reference databases generated by CAD4D, SPM, and NEUROSTAT showed similar AD-related hypometabolism patterns.
• The extent and magnitude of detected hypometabolism were influenced by the software's calculation method, not solely by normalization parameters.
• These findings highlight the importance of considering software-specific calculation methods for reliable dementia diagnosis using FDG-PET.