Abstract
PURPOSE
To investigate if volumetric MRI biomarkers help across both parametric and nonparametric Alzheimer's disease (AD) progression modeling using neuropsychological tests for decline prediction of mini-mental state examination (MMSE) score in converting and stable mild cognitive impairment (MCI) subjects.
METHOD AND MATERIALS
The study dataset consisted of yearly visits (2005-2016) for 372 Alzheimer's Disease Neuroimaging Initiative subjects with normal cognition, MCI, and AD, including the following measurements: FreeSurfer-based T1-weighted brain MRI volumes of ventricles, hippocampus, whole brain, fusiform, middle temporal gyrus, and entorhinal cortex, all normalized with intracranial volume, and cognitive tests of MMSE, CDR-SB, and ADAS-Cog. Two state-of-the-art disease progression modeling methods, a nonparametric [DOI:10.1016/j.media.2019.01.004] and a parametric [DOI:10.1016/j.neurobiolaging.2014.03.043], were trained on the data with and without MRI biomarkers using 336 subjects and were subsequently applied to predict month 24 to 60 MMSE scores for 36 independent test subjects based on only their baseline and month 12 visits.
RESULTS
The predictive power and prognostic capability of the AD progression modeling methods were assessed using the per-visit mean absolute error (MAE) and area under the ROC curve (AUC) of predicted MMSE scores for stable (MCI-to-MCI) and converting (MCI-to-AD) test subjects. The MAE results for month 24 to 60 were as follows: parametric-MRI 1.23 to 4.41 (stable), 1.54 to 11.57 (converting); parametric+MRI 1.09 to 4.39 (stable), 1.72 to 10.98 (converting); nonparametric-MRI 0.93 to 5.27 (stable), 1.62 to 8.28 (converting); nonparametric+MRI 0.23 to 0.46 (stable), 1.63 to 6.79 (converting). The AUC results for month 24 to 60 were as follows (p < 0.01): parametric-MRI 0.90 for all visits; parametric+MRI 0.89 to 0.91; nonparametric-MRI 0.86 to 0.89; nonparametric+MRI 0.85 to 0.95.
CONCLUSION
MRI measurements improve neuropsychological assessment-based disease progression modeling performance of both parametric and non-parametric methods in MMSE decline prediction. Predictions from both utilized methods can significantly discriminate between stable MCI and MCI converting to AD.
Original language | English |
---|---|
Publication date | 2019 |
Number of pages | 1 |
Publication status | Published - 2019 |
Event | RSNA 2019 - 105th Scientific Assembly and Annual Meeting - Chicago, United States Duration: 1 Dec 2019 → 4 Dec 2019 |
Conference
Conference | RSNA 2019 - 105th Scientific Assembly and Annual Meeting |
---|---|
Country/Territory | United States |
City | Chicago |
Period | 01/12/2019 → 04/12/2019 |