Effect of image resolution on the diagnostic performance of disease-related patterns derived from brain FDG-PET images

Dementia is often caused by Alzheimer’s disease (AD) but may have other causes.  Finding the cause may be aided by characteristic metabolic brain patterns exracted from 18F-fluorodeoxyglucose positron emission tomography images. We investigated the effect of image resolution on the AD-related pattern’s (ADRP) diagnostic performance.

A total of 240 subjects (120 AD patients, 120 healthy controls (HC)) were selected from the ADNI database. Images were smoothed to various resolutions. 100 AD and 100 HC were randomly combined into 25 demographically similar subgroups with 20 AD and 20 HC that were used to identify ADRPs. The remaining 20 AD and 20 HC served as a validation group to obtain ROC curves and AUC values.

AUC values ranged from 0.60 to 0.88 and were substantially affected by the group’s subject selection. The best median value of 0.82 was achieved when identification and validation groups had a resolution of 8 mm. Different resolutions in the validation group had almost negligible effect, while the change in resolution for the identification group decreased median AUC for up to 0.02.

Image resolution has some but insignificant effect on ADRPs diagnostic performance. It is also not crucial to have the same image resolution for the identification and validation group.