Optimizing spatial specificity and signal quality in fNIRS: an overview of potential challenges and possible options for improving the reliability of real-time applications

Abstract

The optical brain imaging method functional near-infrared spectroscopy (fNIRS) is a promising tool for real-time applications such as neurofeedback and brain-computer interfaces. Its combination of spatial specificity and mobility makes it particularly attractive for clinical use, both at the bedside and in patients' homes. Despite these advantages, optimizing fNIRS for real-time use requires careful attention to two key aspects: ensuring good spatial specificity and maintaining high signal quality. While fNIRS detects superficial cortical brain regions, consistently and reliably targeting specific regions of interest can be challenging, particularly in studies that require repeated measurements. Variations in cap placement coupled with limited anatomical information may further reduce this accuracy. Furthermore, it is important to maintain good signal quality in real-time contexts to ensure that they reflect the true underlying brain activity. However, fNIRS signals are susceptible to contamination by cerebral and extracerebral systemic noise as well as motion artifacts. Insufficient real-time preprocessing can therefore cause the system to run on noise instead of brain activity. The aim of this review article is to help advance the progress of fNIRS-based real-time applications. It highlights the potential challenges in improving spatial specificity and signal quality, discusses possible options to overcome these challenges, and addresses further considerations relevant to real-time applications. By addressing these topics, the article aims to help improve the planning and execution of future real-time studies, thereby increasing their reliability and repeatability.

Keywords: BCI; extracerebral systemic activity; fNIRS; motion artifacts; neurofeedback; noise reduction; real-time preprocessing; spatial specificity.

Figure 1

Overview of strategies covered to improve spatial specificity. Possible options discussed in this context include probe design, cap placement and spatial validation.

Figure 2

Exemplary probe placement relative to the 10–20 EEG system.

Figure 3

Overview of the strategies covered to improve signal quality in the context of real-time applications. Based on typical offline methods, channel quality assessment and correction, the modified Beer Lambert law, correction of motion artifacts, temporal filtering and correction of extracerebral systemic artifacts are discussed in this context.

Figure 4

Comparison of signal qualities of a poor (A) and a good (D) channel of ΔOD at 760 and 850 nm during resting state data from a single subject. Enlarged signal windows in (B, E) illustrate the presence or absence of a clear cardiac pulsation. The corresponding power spectra resulting from a single-channel fast Fourier transform is shown in (C, F). Quality metrics SCI and PSP are shown in (A, D). Note that ΔOD data is visualized but SCI and PSP are calculated based on the raw light intensity data.

Figure 5

Illustration of MAs (some are black framed) in ΔOD data for wavelengths 760 nm (blue) and 850 nm (red). The data displayed is resting-state data from a single subject.

Figure 6

Visualization of (A) Δ[HbX] signals including MAs, (B) corresponding 3D accelerometer data and (C) 3D gyroscope data. The data displayed is motor execution data from a single subject.

Figure 7

Illustration of the effect of offline MA correction. (A) shows only band-pass filtered Δ[HbX] data without MA correction, (B) Δ[HbX] data corrected with TDDR + band-pass filtering and (C) Δ[HbX] data corrected with CBSI + band-pass filtering. Applied filter was a zero-phase 2nd order Butterworth filter with cut-off frequencies of [0.01, 0.09] Hz and all preprocessing steps were performed offline. The data displayed is motor execution data from a single subject.

Figure 8

Illustration of the effects of different filter types applied to Δ[HbO] (red) and Δ[HbR] (blue) of semi-simulated data based on resting state data from a single subject. (A) shows the unfiltered signals. (B) Demonstrates the result of a low-pass filter with f_tlow = 0.09 Hz, effectively smoothing the signal. (C) shows the outcome of a high-pass filter with f_thigh = 0.01 Hz, effectively removing slow drifts. (D) shows the output of a band-pass filter with a cut-off frequency range of [0.01, 0.09] Hz. The filters were applied offline using a zero-phase 2nd order Butterworth filter. The gray areas indicate task periods.

Figure 9

Illustration of the effect of causal and acausal filters for (A) IIR and (B) FIR filters. IIR: filter order = 2, Butterworth; FIR: filter order = 450. Note that band-pass filters with [0.01, 0.09] Hz were applied offline. The gray areas indicate task periods. The data displayed is semi-simulated data based on resting state data from a single subject.

Figure 10

Comparison of offline filtered data with different cut-off frequencies. (A) Data filtered with cut-offs of [0.01, 0.5] Hz and (B) data filtered with cut-offs of [0.01, 0.09] Hz. The vertical black line indicates the same time point in both figures. The data displayed is semi-simulated data based on resting state data from a single subject.

Figure 11

Illustration of the penetration depth of a short-distance channel (0.8 cm source-detector distance) as compared to a regular-distance channel (~3 cm source-detector distance).

Figure 12

Illustration of the effects of extracerebral systemic artifact correction methods with and without SDCs on two subjects (A, B) s01, and (C, D) s02 that performed a motor execution task. The task-evoked time series are shown for preprocessed data without correction (UNCORRECTED). Correction methods without SDCs, including CAR, PCA, and GCR, as well as correction methods with SDCs, such as GLM-filter-based correction using eight SDCs (GLM FILTER) and SSR with the spatially closest SDC, are used to present the data corrected for extracerebral systemic artifacts. Gray areas indicate task periods.

Figure 13

Tabular overview of all options discussed in this paper that could help improve spatial specificity, including potential challenges, possible options and further considerations. Please note that this overview does not claim to be complete and only represents suggestions.

Figure 14

Tabular overview of all preprocessing steps discussed in this review that might help improve signal quality in real-time applications, including potential challenges, possible options and further considerations. Please note that this overview does not claim to be complete and only represents suggestions.