Abstract
1455
Introduction: PET/MR systems are increasingly used for clinical neurodegenerative evaluation and research. One paradigmatic PET/MR application is the assessment of dementia. However, suboptimal MR-based attenuation correction (MRAC) on clinical PET/MR causes quantification errors in PET images, which may degrade diagnostic accuracy. The purpose of this study was to assess the impact of clinical MRAC on the evaluation of Alzheimer’s disease (AD).
Methods: We recruited 47 patients from two institutions who underwent PET/CT and PET/MR (GE SIGNA) examination for oncological staging. From the PET raw data acquired on PET/MR, two FDG-PET series were generated, based on clinical MRAC (atlas-based method) and CTAC. The following simulation steps were performed in MNI space: After spatial normalization and smoothing of the PET datasets, we calculated the error map for each patient, as PET based on MRAC divided by that based on CTAC. We multiplied each of these 47 error maps with each of the 203 Alzheimer’s Disease Neuroimaging Initiative (ADNI) cases after the identical normalization and smoothing. This resulted in 203[asterisk]47= 9541 datasets. To assess the diagnostic accuracy of the resulting images in AD, an AD PET score was calculated automatically by using a standard software (PMOD PALZ) which incorporates a method published by Herholz et al. (2002).
Results: The accuracy, specificity for the discrimination of AD-patients from normal controls was not substantially impaired but sensitivity was slightly impaired in 5 out of 47 dataset (original vs. error; 83.2% [CI 75.0%-89.0%], 83.3% [CI 74.2%-89.8%] and 83.1% [CI 75.6%-88.3%] vs. 82.7% [range 80.4-85.0%], 78.5% [range 72.9-83.3%,] and 86.1% [range 81.4-89.8%]). The accuracy, sensitivity and specificity for predicting progression from MCI to AD during 2-year follow-up was not impaired in any of the 47 dataset (original vs. error; 62.5% [CI 53.3%-69.3%], 78.8% [CI 65.4%-88.6%] and 54.0% [CI 47.0%-69.1%] vs. 64.8% [range 61.5-66.7%], 75.7% [range 66.7-81.8%,] and 59.0% [range 50.8-63.5%]). The worst 3 error maps show a tendency towards underestimation of PET scores.
Conclusions: Clinical atlas-based MR attenuation correction is expected to have sufficient diagnostic accuracy for the diagnosis of Alzheimer’s disease and for the prediction of mild cognitive impairment progression to Alzheimer’s disease, although sensitivity is slightly impaired.