Fuzzy cluster analysis of high-field functional MRI data

Area of Science:

• Neuroimaging
• Data Analysis
• Biophysics

Background:

• Functional magnetic resonance imaging (fMRI) using blood-oxygen-level dependent (BOLD) contrast is a key brain research tool, but understanding its mechanisms and artifacts remains challenging.
• Exploratory data analysis (EDA) is crucial for modeling fMRI signal changes, identifying artifacts, and quantifying neural activity.
• Very high-field fMRI presents unique opportunities and challenges for data analysis.

Purpose of the Study:

• To investigate the utility of fuzzy cluster analysis (FCA) for identifying and separating artifacts in fMRI data.
• To differentiate between true neural activation and false vascular signals in fMRI.
• To explore the application and limitations of FCA in very high-field fMRI.

Main Methods:

• Applied fuzzy cluster analysis (FCA) to fMRI time series data, including synthetic and in vivo datasets.
• Utilized a test object with static and dynamic parts to differentiate temporal patterns and head motion artifacts.
• Quantitatively evaluated FCA parameters using receiver-operator characteristics (ROC) and compared with correlation analysis (CA).

Main Results:

• FCA effectively identified and separated various artifacts in fMRI data, including those from head motion.
• Demonstrated the ability to differentiate true neural activation from false vascular signals based on echo time dependence and signal time-course.
• FCA provided novel insights into fMRI data interpretation, classification, and characterization.

Conclusions:

• Fuzzy cluster analysis is a valuable exploratory data analysis tool for improving the reliability and interpretation of fMRI studies.
• FCA can enhance the design of future fMRI acquisition schemes, paradigms, and biophysical models.
• The method aids in distinguishing neural from vascular signals, crucial for clinical diagnosis and neuroscience research.