Related Experiment Video
Updated: Sep 11, 2025
Hybrid PET/MRI Imaging of Alzheimer's Disease Based on 18F-AV-1451
Published on: April 18, 2025
Classifying mild cognitive impairment from normal cognition: fMRI complexity matches tau PET performance
Kay Jann1, Gilsoon Park1, John M Ringman2
1USC Stevens Neuroimaging and Informatics Institute Keck School of Medicine University of Southern California Los Angeles California USA.
Introduction:
Cognitive decline in Alzheimer's disease (AD) is closely linked to tau pathology, which leads to loss of synaptic connections and ultimately neurons. While tau positron emission tomography (PET) carries radiation risks, is costly, and often unavailable in clinical settings, brain entropy mapping via resting-state functional magnetic resonance imaging (fMRI) has emerged as a marker of impaired brain function related to tauopathy.
Methods:
Using data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and Estudio de la Enfermedad de Alzheimer en Jalisciences (EEAJ), we investigate the classification performance of fMRI entropy with tau PET in distinguishing cognitively normal (CN) from cognitively impaired (mild cognitive impairment/AD) individuals. Convolutional neural networks, initially trained and evaluated via 5-fold cross-validation on ADNI data, were subsequently tested on an independent external cohort (EEAJ) using an ensemble approach.
Results:
The fMRI entropy classifier matched the tau PET model in accuracy and outperformed it in F1 score (0.64 vs. 0.61) and area under the curve (AUC; 0.73 vs. 0.67). On the independent external validation dataset (EEAJ), fMRI sample entropy showed a comparable F1 score (0.88) to tau PET (0.88) and achieved a notably higher AUC (0.94 vs. 0.92).
Discussion:
Our findings suggest that fMRI entropy could be a non-invasive imaging marker alternative to tau PET for detecting AD-related cognitive impairment.
Highlights:
Functional magnetic resonance imaging (fMRI) complexity matches tau positron emission tomography (PET) in classifying cognitive impairment.Sample entropy and multiscale entropy were used for fMRI-based Alzheimer's disease (AD) classification.3D convolutional neural networks models achieve up to 84% accuracy using fMRI complexity measures.The dorsal attention network was identified as critical for distinguishing mild cognitive impairment/AD.fMRI complexity offers a non-invasive alternative to tau positron emission tomography imaging.
