Format

Send to

Choose Destination
Eur J Nucl Med Mol Imaging. 2015 May;42(6):916-27. doi: 10.1007/s00259-015-2994-9. Epub 2015 Feb 3.

Visual and statistical analysis of ¹⁸F-FDG PET in primary progressive aphasia.

Author information

1
Department of Neurology, Hospital Clínico San Carlos, Madrid, Spain, jordimatiasguiu@hotmail.com.

Abstract

PURPOSE:

Diagnosing progressive primary aphasia (PPA) and its variants is of great clinical importance, and fluorodeoxyglucose (FDG) positron emission tomography (PET) may be a useful diagnostic technique. The purpose of this study was to evaluate interobserver variability in the interpretation of FDG PET images in PPA as well as the diagnostic sensitivity and specificity of the technique. We also aimed to compare visual and statistical analyses of these images.

METHODS:

There were 10 raters who analysed 44 FDG PET scans from 33 PPA patients and 11 controls. Five raters analysed the images visually, while the other five used maps created using Statistical Parametric Mapping software. Two spatial normalization procedures were performed: global mean normalization and cerebellar normalization. Clinical diagnosis was considered the gold standard.

RESULTS:

Inter-rater concordance was moderate for visual analysis (Fleiss' kappa 0.568) and substantial for statistical analysis (kappa 0.756-0.881). Agreement was good for all three variants of PPA except for the nonfluent/agrammatic variant studied with visual analysis. The sensitivity and specificity of each rater's diagnosis of PPA was high, averaging 87.8 and 89.9% for visual analysis and 96.9 and 90.9% for statistical analysis using global mean normalization, respectively. In cerebellar normalization, sensitivity was 88.9% and specificity 100%.

CONCLUSION:

FDG PET demonstrated high diagnostic accuracy for the diagnosis of PPA and its variants. Inter-rater concordance was higher for statistical analysis, especially for the nonfluent/agrammatic variant. These data support the use of FDG PET to evaluate patients with PPA and show that statistical analysis methods are particularly useful for identifying the nonfluent/agrammatic variant of PPA.

PMID:
25647075
DOI:
10.1007/s00259-015-2994-9
[Indexed for MEDLINE]

Supplemental Content

Full text links

Icon for Springer
Loading ...
Support Center