The impact of partial volume effect correction on brain network analysis using AV-45 PET data.

Objectives The quantitative accuracy of PET is degraded by partial volume effects caused by limited spatial resolution of PET scanners. The goal of this study was to evaluate the effectiveness of partial volume correction (PVC) [1] on AV-45 PET image and consequently its impact on brain network analysis. Brain network at different stages of Alzheimer’s disease (AD) was constructed to demonstrate of the impact of PVC.

Methods The AV-45 PET scans of 416 subjects in four different stages of AD, including NC (Normal Control), EMCI (Early Mild Cognitive Impairment), LMCI (Late Mild Cognitive Impairment) and AD, in ADNI (Alzheimer’s Disease Neuroimaging Initiative) database were analyzed. After image registration to MNI-152 brain template, PVC was performed on all images. Then brain networks of four stages of AD were generated from both pre- and post-partial volume corrected images. For each stage and each group (pre and post PVC) we then randomly selected 90% subjects to generate brain network and repeated 200 times to obtain 200 brain networks of each group. We performed permutation test on these 200 networks and did t-test on all different edges within stages.

Results Spatial resolution of PET images was improved after PVC: the boundary of gray matter and white matter was strengthened and the details became more visible. Mean network degrees of LMCI group decreased significantly (from 11.0769±7.9717 to 8.4423±5.7502, network absolute threshold rate r=0.8) while other groups remained almost the same. The strength of nodes of brain networks also changed after PVC, especially in LMCI group, where 95 of total 104 regions decreased and 30.53% of them decreased more than 10%. Mean node strength of LMCI group decreased from 59.5318±8.2607 to 55.6213±8.4887 after PVC. 59 more edges (22.61% of insignificant edges before PVC) of whole brain network in LMCI and AD groups showed significant difference (p=0.05) after PVC, 118 more edges (47.98%) in EMCI and LMCI groups and 96 more edges (76.19%) in EMCI and AD groups, which suggests the effectiveness of PVC in distinguishing EMCI, LMCI and AD patients.

Conclusions Partial volume correction can improve the quantitative accuracy of AV-45 PET and consequently the accuracy of brain network analysis, especially in distinguishing EMCI, LMCI and AD patients.$$graphic_{E39B9FCF-A106-410A-BB22-CD3F11EB848F}$$

• Download figure
• Open in new tab
• Download powerpoint