Bootstrap resampling method to estimate confidence intervals of activation-induced CBF changes using laser Doppler...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Laser Doppler imaging (LDI) signal and noise characteristics vary with vascular caliber and are non-stationary.
• Conventional statistical methods assume constant noise and normal distribution, leading to inaccurate activation detection.
• Concatenating experimental data can introduce temporal noise variations.

Purpose of the Study:

• To apply bootstrap resampling with cross-correlation for pixel-by-pixel confidence interval estimation in LDI.
• To improve the detection of whisker activation-induced cerebral blood flow (CBF) changes.
• To overcome limitations of traditional methods in non-stationary noise environments.

Main Methods:

• Utilized bootstrap resampling in conjunction with cross-correlation analysis.
• Estimated confidence intervals on a pixel-by-pixel basis.
• Applied to analyze whisker activation-induced CBF changes in the cerebral cortex.

Main Results:

• Bootstrap resampling increased the number of detected active pixels by approximately 45% at a 95% confidence level compared to conventional cross-correlation.
• The detected pixels were predominantly in regions with intermediate and large baseline LDI flux and significant deviation from normality.
• This method proved effective in regions with high temporal noise variation and low contrast-to-noise ratio (CNR).

Conclusions:

• Bootstrap-based confidence intervals provide unbiased detection of CBF changes.
• This approach is particularly valuable for LDI data with non-stationary noise and non-normal distributions.
• Enhances the reliability of neurovascular coupling studies using LDI.