Statistical framework and noise sensitivity of the amplitude radial correlation contrast method

Area of Science:

• Neuroimaging
• Statistical analysis
• Functional magnetic resonance imaging (fMRI)

Background:

• The radial correlation contrast (RCC) method analyzes functional MRI (fMRI) data by assessing temporal cross-correlation between neighboring voxels.
• Comparing brain states (e.g., stimulation ON vs. OFF) is crucial for understanding neural activity.
• Existing methods may benefit from enhanced statistical rigor for improved accuracy.

Purpose of the Study:

• To present a novel statistical framework for the amplitude RCC method.
• To integrate conventional pixel thresholding with cluster-size statistics for fMRI analysis.
• To establish robust cutoffs for identifying significant brain activity changes.

Main Methods:

• Developed a statistical framework combining pixel thresholding and cluster-size statistics for RCC analysis.
• Defined RCC correlation maps by differencing RCC images from different brain states.
• Determined pixel and cluster-size cutoffs using normal distribution properties and empirical null distributions to control false positives.
• Derived an analytical expression for amplitude-RCC dependency on noise to set the pixel threshold.
• Utilized in vivo and ex vivo rat fMRI data during forepaw stimulation for threshold fine-tuning.

Main Results:

• The proposed framework establishes a normal distribution for the RCC map in OFF states, enabling pixel cutoff definition.
• An empirical cluster-size null distribution was used to set a cluster-size cutoff, allowing for 5% false positives.
• An analytical expression relating amplitude-RCC to noise was derived and used to define the pixel threshold.
• In vivo and ex vivo data analysis showed enhanced spatial coherence in in vivo images.
• The proposed cutoffs demonstrated generality, independent of anesthesia method, magnetic field strength, or anesthesia depth.

Conclusions:

• The presented statistical framework enhances the amplitude RCC method for fMRI analysis.
• The developed cutoffs provide a statistically sound basis for identifying significant changes in brain activity.
• The findings support the general applicability of the proposed methodology across various experimental conditions.