Evaluation of sliding window correlation performance for characterizing dynamic functional connectivity and brain
Sadia Shakil1, Chin-Hui Lee1, Shella Dawn Keilholz2
1Georgia Institute of Technology, Electrical and Computer Engineering, Atlanta, GA, USA.
Neuroimage
|March 9, 2016
Summary
Sliding window correlation (SWC) analysis of resting-state functional MRI data is sensitive to parameter choices. Optimal window length is crucial for accurately detecting brain state transitions, suggesting a need for adaptive methods.
Area of Science:
- Neuroimaging
- Computational Neuroscience
- Brain Dynamics
Background:
- Resting-state functional MRI (fMRI) analysis is vital for understanding brain function and disorders.
- Dynamic analysis of fMRI data offers insights into cognitive processes and neuropathology.
- Sliding window correlation (SWC) is a common method for assessing dynamic functional connectivity.
Purpose of the Study:
- To evaluate the performance of sliding window correlation (SWC) in analyzing resting-state fMRI data.
- To investigate the impact of various parameters on SWC's ability to detect brain state transitions.
- To assess the reliability of SWC-based clustering for identifying network configurations.
Main Methods:
- Utilized simulated neural networks (SNs) with predefined state transitions.
- Systematically varied parameters including window length, offset, type, noise, filtering, and sampling rate.
- Calculated SWC time courses, clustered results using k-means, and compared to known SN states.
Main Results:
- Window length and offset significantly influenced the detection of state transitions and durations.
- Noise and filtering parameters also impacted SWC performance; sampling rate had minimal effect.
- Tapered windows showed lower sensitivity to sharp transitions compared to rectangular windows.
- SWC provided suboptimal correlation estimates for brain states; clustering reliability depended on window length matching state duration.
Conclusions:
- SWC performance in dynamic fMRI analysis is highly parameter-dependent, particularly window length.
- Current SWC methods may not reliably capture underlying brain state dynamics.
- Development of adaptive windowing techniques is necessary for accurate resting-state network analysis.
Keywords:
Functional connectivityNetwork dynamicsResting-state functional MRISliding window correlationStatesk-Means
