Characterizing near-infrared spectroscopy signal under hypercapnia

Abstract

Vasoactive stress tests (i.e. hypercapnia, elevated partial pressure of arterial CO₂ [PaCO₂ ]) are commonly used in functional MRI (fMRI), to induce cerebral blood flow changes and expose hidden perfusion deficits in the brain. Compared with fMRI, near-infrared spectroscopy (NIRS) is an alternative low-cost, real-time, and non-invasive tool, which can be applied in out-of-hospital settings. To develop and optimize vasoactive stress tests for NIRS, several hypercapnia-induced tasks were tested using concurrent-NIRS/fMRI on healthy subjects. The results indicated that the cerebral and extracerebral reactivity to elevated PaCO₂ depended on the rate of the CO₂ increase. A steep increase resulted in different cerebral and extracerebral reactivities, leading to unpredictable NIRS measurements compared with fMRI. However, a ramped increase, induced by ramped-CO₂ inhalation or breath-holding tasks, induced synchronized cerebral, and extracerebral reactivities, resulting in consistent NIRS and fMRI measurements. These results demonstrate that only tasks that increase PaCO₂ gradually can produce reliable NIRS results.

Keywords: cerebral blood flow; hypercapnia; magnetic resonance imaging; near-infrared spectroscopy.

Fig. 1.

Experimental design and setup. Schematic of (a) the sharp-CI, (b) the long-ramped CI, (c) the BH, and (d) the short-ramped CI tasks. (e) Configuration of the NIRS channels on the head. (f) Experimental setup of the concurrent fMRI/NIRS experiments.

Fig. 2.

Averaged MCCC from each channel under the sharp-CI task between $\text{Δ}\overline{BOLD}\%$ and (a) Δ[HbO], (b) Δ[Hb]. Crosses in the channel indicate a p-value larger than 0.05 under the FDR-criterion.

Fig. 3.

Results showing inconsistencies in NIRS signals during the sharp-CI task from two example subjects in (a) and (b). The signals of ΔP_ETCO₂, $\text{Δ}\overline{BOLD}\%$, Δ[HbO], and Δ[Hb] are shown in the colors of black, purple, red, and blue respectively. The averaged results of Δ[HbO] and Δ[Hb]are shown at bottom right of (a) and (b) in red boxes.

Fig. 4.

The fMRI and NIRS results from the (a) BH, (b) long-ramped CI, (c) short-ramped CI, and (d) sharp-CI tasks. The shaded areas indicate the targeted ΔP_ETCO₂ in each task. The averaged ΔP_ETCO₂, $\text{Δ}\overline{BOLD}\%$ (normalized), Δ[HbO], and Δ[Hb] signals are shown in columns 1-4, respectively.

Fig. 5.

Averaged MCCCs between $\text{Δ}\overline{BOLD}\%$ and Δ[HbO] (Δ[Hb]) signals are shown in column 1 (2). The corresponding tasks are: BH (a), long-ramped CI (b), short-ramped CI (c), sharp-CI (d). Crosses in the channel indicate a p-value larger than 0.05 under one-sample t-test.

Fig. 6.

The fMRI movie snapshots from one subject under the long-ramped CI task. Three snapshots were captured at (1) rising phase (2) plateau, and (3) falling phase of ΔP_ETCO₂ in the long-ramped CI task. During each time point, real-time (a) ΔP_ETCO₂ and (b) $\text{Δ}\overline{BOLD}\%$ (normalized) signals are displayed. Also, $\text{Δ}\overline{BOLD}\%$ (normalized) maps (red-yellow refers to positive percent changes and blue-light blue refers to negative percent changes) in (c) sagittal, (d) coronal, and (e) axial views are shown. The shaded areas indicate the targeted ΔP_ETCO₂ during the task. The black square in every (c) and (e) panel indicate the region of prefrontal cortex (PFC).

Fig. 7.

Group folding average of signals from the (a) BH and (b) short-ramped CI task. The results of $\text{Δ}\overline{BOLD}\%$ (normalized), Δ[HbO], and Δ[Hb] are showed in columns 1-3, respectively. The shaded areas (prior to 20 s) indicate the periods of BH/ramped-increased ΔP_ETCO₂.