Resting-state multi-spectrum functional connectivity networks for identification of MCI patients

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Mild cognitive impairment (MCI) is a precursor to Alzheimer's disease, necessitating accurate early detection methods.
• Resting-state functional magnetic resonance imaging (fMRI) reveals brain functional connectivity patterns.
• Blood oxygenation level dependent (BOLD) signals in fMRI exhibit frequency-specific properties.

Purpose of the Study:

• To develop a high-dimensional pattern classification framework for accurately identifying individuals with MCI from healthy controls.
• To leverage multi-spectrum functional brain networks to characterize subtle BOLD signal changes associated with MCI.
• To enhance the early detection of functional brain abnormalities for improved management of potential Alzheimer's disease.

Main Methods:

• A novel framework using multi-spectrum networks based on functional associations between brain regions during resting-state.
• Regional time series were band-pass filtered into five frequency sub-bands, creating five distinct connectivity networks.
• Clustering coefficients of regions of interest (ROIs) were extracted as classification features.
• Leave-one-out cross-validation was employed to evaluate classification accuracy and generalization performance.

Main Results:

• The proposed multi-spectrum network approach achieved a classification accuracy of 86.5% for MCI detection.
• This represents an improvement of at least 18.9% compared to conventional full-spectrum methods.
• The area under the receiver operating characteristic (ROC) curve was estimated at 0.863, indicating good diagnostic power.
• Specific brain regions, including parts of the prefrontal cortex, orbitofrontal cortex, temporal lobe, and parietal lobe, showed the most discriminant information.

Conclusions:

• The multi-spectrum network framework offers a significant advancement in the early and accurate detection of MCI.
• Frequency-specific analysis of BOLD signals provides more effective characterization of subtle pathological changes.
• This approach contributes positively to the treatment management of potential Alzheimer's disease patients by enabling earlier intervention.