Identification of MCI individuals using structural and functional connectivity networks

Area of Science:

• Neuroimaging
• Brain network analysis
• Neurological disorder classification

Background:

• Mild cognitive impairment (MCI) is an early stage of Alzheimer's disease (AD), often presenting with subtle cognitive symptoms, making early diagnosis challenging.
• Brain network analysis offers novel methods for characterizing neurological disorders by examining whole-brain connectivity.

Purpose of the Study:

• To integrate multiple imaging modalities for enhanced identification of individuals at risk for MCI.
• To improve classification performance for MCI detection by combining diffusion tensor imaging (DTI) and resting-state functional magnetic resonance imaging (rs-fMRI).

Main Methods:

• Utilized multiple-kernel Support Vector Machines (SVMs) to integrate data from DTI and rs-fMRI.
• Compared multimodality classification accuracy against single-modality methods and direct data fusion.

Main Results:

• The multimodality approach achieved a classification accuracy of 96.3%, a statistically significant improvement over single modalities.
• Accuracy increased by at least 7.4% compared to single-modality and direct data fusion methods.
• Cross-validation yielded an area of 0.953 under the receiver operating characteristic (ROC) curve, demonstrating excellent diagnostic power.

Conclusions:

• Integrating DTI and rs-fMRI through multimodality classification significantly enhances the accuracy of early MCI detection.
• This approach offers greater sensitivity in identifying brain abnormalities, aiding in the early diagnosis of neurodegenerative conditions like AD.