Reducing false discoveries in resting-state functional connectivity using short channel correction: an fNIRS study

Abstract

Significance: Functional near-infrared spectroscopy (fNIRS) is a neuroimaging tool that can measure resting-state functional connectivity; however, non-neuronal components present in fNIRS signals introduce false discoveries in connectivity, which can impact interpretation of functional networks. Aim: We investigated the effect of short channel correction on resting-state connectivity by removing non-neuronal signals from fNIRS long channel data. We hypothesized that false discoveries in connectivity can be reduced, hence improving the discriminability of functional networks of known, different connectivity strengths. Approach: A principal component analysis-based short channel correction technique was applied to resting-state data of 10 healthy adult subjects. Connectivity was analyzed using magnitude-squared coherence of channel pairs in connectivity groups of homologous and control brain regions, which are known to differ in connectivity. Results: By removing non-neuronal components using short channel correction, significant reduction of coherence was observed for oxy-hemoglobin concentration changes in frequency bands associated with resting-state connectivity that overlap with the Mayer wave frequencies. The results showed that short channel correction reduced spurious correlations in connectivity measures and improved the discriminability between homologous and control groups. Conclusions: Resting-state functional connectivity analysis with short channel correction performs better than without correction in its ability to distinguish functional networks with distinct connectivity characteristics.

Keywords: functional near-infrared spectroscopy; magnitude-squared coherence; physiological noise removal; principal component analysis; resting-state functional connectivity; short channel correction.

© 2022 The Authors.

Figures

Fig. 1

Illustration of the optode montage and the connectivity groups used in this study. (a) Schematic diagram showing the differences in light path and penetration depth of long channels and short channels. (b) Top-view of the optode montage. Sources and detectors are marked with red and brown dots and their numbers are displayed in red and blue circles, respectively. Channels are marked with solid white lines and the mid-point of each channel is marked with a yellow dot. (c) Side-view of the optode montage. Registered channel positions are shown on left and right hemispheres of the brain with respect to the landmarks of international 10–20 standard (i.e., Nz, Cz, Iz, LPA, and RPA). (d) Physical layout of dual-tip optodes. (e) Channel pair definition of the connectivity groups; homologous connectivity group linking channels in interhemispheric frontal, temporal, and occipital homologous regions and control connectivity group comprises long distance connections that have no known direct structural linkage to date.

Fig. 2

Representative example of resting state signals and the effect of physiological noise on resting-state functional connectivity (a) Configuration of the optode montage with 10 sources (red dots), eight detectors (blue dots) and 16 long distance channels (solid gray lines). A representative homologous channel pair (marked in black and green) and a control channel pair (marked in black and pink) are shown in the montage. (b) Sample time traces of resting-state HbO concentration changes of subject #3. The time series signals are color-coded with the channel colors of Fig. 2(a). (c) The power spectrum of the three signals depicting the presence of systemic physiological noise in fNIRS signals such as Mayer waves (∼0.1  Hz), respiration (0.2 to 0.3 Hz) and heartbeat (∼1  Hz). The partial overlap between the Mayer wave frequencies (M-band: 0.05 to 0.15 Hz) and the frequency band of interest for resting-state functional connectivity (RSFC-band: 0.009 to 0.1 Hz) is also highlighted.

Fig. 3

Signal processing pipeline of the resting-state fNIRS data. (a) Common pipeline was used until the raw fNIRS signals were converted to concentration changes of HbO and HbR. Functional connectivity analysis was performed without and with short channel correction to evaluate the effect of removing systemic physiological noise on resting-state functional connectivity measures (demarcated by shaded blue box). (b) Important parameters used in each step of the pipeline. (c) Representative example of a channel affected by step-like motion artefacts (subject #2, channel: S10-D8, time points of motion artefacts marked by gray circles) are shown with plots of optical density, motion artefact corrected: with TDDR, with wavelet filter for signals of two wavelengths (760 and 850 nm) and HbO and HbR concentration changes.

Fig. 4

Signal quality assessment of long channels for each subject. (a) Each dot represents scalp coupling index of a long channel for each subject. The gray dashed line denotes the threshold of the scalp coupling index (SCI=0.5) used in this study. Channels that have SCI below the threshold were excluded from the connectivity analysis. (b) The bar chart shows the number of channel pairs retained in the connectivity analysis for each subject in homologous and control groups.

Fig. 5

Comparison of magnitude-squared coherence of HbO signals across four contrasting conditions (a) Channel pair definition of the connectivity groups. Homologous and control group contains interhemispheric bilateral channel pairs and channel pairs with no known direct structural linkage, respectively. (b) Coherence plots for intra-subject homologous, intra-subject control, inter-subject random and white Gaussian noise (baseline) conditions without (left) and with short channel correction (right). Dashed vertical gray lines indicate resting-state frequency band (0.009 to 0.1 Hz). Mayer wave frequency band (0.05 to 0.15 Hz), respiratory band (0.2 to 0.3 Hz) and heartbeat band (∼1  Hz) are highlighted in shaded gray boxes with increasing color intensity.

Fig. 6

Participant-averaged magnitude-squared coherence results of long channel and short channel pairs of homologous groups. (a) Channel pair definition of homologous long channels and homologous short channels. (b) Coherence plots of HbO signals (left) and HbR signals (right) for short channel data and long channel data with and without short channel correction. Long channel coherence without correction and short channel coherence show similar peaks for both HbO and HbR signals around the frequencies associated with physiological noise such as Mayer waves, respiration, and heartbeat at ∼0.1, ∼0.2 and ∼1  Hz, respectively.

Fig. 7

Participant-averaged magnitude-squared coherence results for two connectivity groups with and without short channel correction. (a) Comparison of coherence of HbO signals without (left) and with short channel correction (right). (b) Comparison of coherence of HbR signals without (left) and with short channel correction (right). (c) Bar chart represents mean coherence of frequency band 0.05 to 0.1 Hz for both connectivity groups with and without short channel correction for HbO (left) and HbR (right). Error bars represent the standard deviation of coherence across subjects.