Affiliations: [a] Department of Nuclear Medicine, Charité - Universitätsmedizin Berlin, Berlin, Germany
| [b] jung diagnostics GmbH, Hamburg, Germany
| [c] Department of Nuclear Medicine, University of Freiburg, Freiburg, Germany
Correspondence:
[*]
Correspondence to: Ralph Buchert, Department of Nuclear Medicine, Charité - Universitätsmedizin Berlin, Berlin, 10117,Germany. Tel.: +49 30 450627059; Fax: +49 30 450527912; E-mail: ralph.buchert@charite.de
Note: [1] Data used in preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (http://adni.loni.usc.edu). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in analysis or writing of this report. A complete listing of ADNI investigators can be found at: http://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf
Abstract: Background:Positron emission tomography (PET) with the glucose analog F-18-fluorodeoxyglucose (FDG) is widely used in the diagnosis of neurodegenerative diseases. Guidelines recommend voxel-based statistical testing to support visual evaluation of the PET images. However, the performance of voxel-based testing strongly depends on each single preprocessing step involved. Objective:To optimize the processing pipeline of voxel-based testing for the prognosis of dementia in subjects with amnestic mild cognitive impairment (MCI). Methods:The study included 108 ADNI MCI subjects grouped as ‘stable MCI’ (n = 77) or ‘MCI-to-AD converter’ according to their diagnostic trajectory over 3 years. Thirty-two ADNI normals served as controls. Voxel-based testing was performed with the statistical parametric mapping software (SPM8) starting with default settings. The following modifications were added step-by-step: (i) motion correction, (ii) custom-made FDG template, (iii) different reference regions for intensity scaling, and (iv) smoothing was varied between 8 and 18 mm. The t-sum score for hypometabolism within a predefined AD mask was compared between the different settings using receiver operating characteristic (ROC) analysis with respect to differentiation between ‘stable MCI’ and ‘MCI-to-AD converter’. The area (AUC) under the ROC curve was used as performance measure. Results:The default setting provided an AUC of 0.728. The modifications of the processing pipeline improved the AUC up to 0.832 (p = 0.046). Improvement of the AUC was confirmed in an independent validation sample of 241 ADNI MCI subjects (p = 0.048). Conclusion:The prognostic value of voxel-based single subject analysis of brain FDG PET in MCI subjects can be improved considerably by optimizing the processing pipeline.
