An analysis of signal processing algorithm performance for cortical intrinsic optical signal imaging and strategies for algorithm selection

M. Nilsson,K. Zalewska,F. R. Walker,J. A. Turley,S. J. Johnson

doi:10.1038/s41598-017-06864-y

Abstract

Intrinsic Optical Signal (IOS) imaging has been used extensively to examine activity-related changes within the cerebral cortex. A significant technical challenge with IOS imaging is the presence of large noise, artefact components and periodic interference. Signal processing is therefore important in obtaining quality IOS imaging results. Several signal processing techniques have been deployed, however, the performance of these approaches for IOS imaging has never been directly compared. The current study aims to compare signal processing techniques that can be used when quantifying stimuli-response IOS imaging data. Data were gathered from the somatosensory cortex of mice following piezoelectric stimulation of the hindlimb. The effectiveness of each technique to remove noise and extract the IOS signal was compared for both spatial and temporal responses. Careful analysis of the advantages and disadvantages of each method were carried out to inform the choice of signal processing for IOS imaging. We conclude that spatial Gaussian filtering is the most effective choices for improving the spatial IOS response, whilst temporal low pass and bandpass filtering produce the best results for producing temporal responses when periodic stimuli are an option. Global signal regression and truncated difference also work well and do not require periodic stimuli.

Highlights

Optical imaging techniques have proven to be a useful tool in understanding changes in brain function
When the brain is stimulated by an external source, for example, tapping on the hind paw of a mouse, a haemodynamic response is triggered, in this case, within the somatosensory cortex[7]
This is typically done with a ring of timed light emitting diodes (LEDs)[17, 18] or a filter wheel[19, 20] with different wavelengths

Summary

Introduction

Optical imaging techniques have proven to be a useful tool in understanding changes in brain function. Selecting an isobestic point, such as 530 nm, weights the changes for both HbO2 and HbR allowing visualisation of the total change in haemoglobin (HbT)[13] This process can be taken a step further by imaging whilst illuminating with a larger number of wavelengths separately, and combining the data using least squares and the modified Beer-Lambert law[14]. This produces a space-time matrix containing the absolute changes in concentrations of HbR and HbO2. The collected imaging data will have a single wavelength for each frame

Methods

Results

Conclusion