Frequency analysis of oscillations in cerebral hemodynamics measured by time domain near infrared spectroscopy

Abstract

In this paper, we propose the application of time-domain near-infrared spectroscopy to the assessment of oscillations in cerebral hemodynamics. These oscillations were observed in the statistical moments of the distributions of time of flight of photons (DTOFs) measured on the head. We analyzed the zeroth and second centralized moments of DTOFs (total number of photons and variance) to obtain their spectra to provide parameters for the frequency components of microcirculation, which differ between the extracerebral and intracerebral layers of the head. Analysis of these moments revealed statistically significant differences between a control group of healthy subjects and a group of patients with severe neurovascular disorders, which is a promising result for the assessment of cerebral microcirculation and cerebral autoregulation mechanisms.

Fig. 1

Two-channel tdNIRS system equipped with four independent pulsed laser sources and two hybrid photomultipliers.

Fig. 2

Representative long-term recording of the changes in statistical moments of the DTOFs obtained during the measurement of a healthy subject at rest at two wavelengths: 687 and 830 nm. Frequency components related to heart rate and respiration are clearly visible in the light attenuation signal (a), but are not clear in the signals for change in the mean time of flight of photons (b) and variance of DTOF (c), where noise contamination is apparently higher.

Fig. 3

Illustrative example of the amplitude spectrum analysis of the ΔA (a) and ΔV (b) signals for a healthy subject. In both graphs, the arterial pulsation components are visible, whereas the respiration arterial pulsation first harmonic and Mayer oscillations are only present in the ΔA (a) amplitude spectrum.

Fig. 4

Illustrative example of the amplitude spectrum analysis of the ΔA (a) and ΔV (b) signals for a TBI patient. In the ΔA signal (a), the arterial pulsation and respiration components are visible, whereas in the variance amplitude spectrum (b), all components are covered by noise, except for the LF components (f < 0.3 Hz).

Fig. 5

Comparison of the amplitude spectra of the light attenuation signal (ΔA) (a and c) and variance (ΔV) (b and d) of DTOFs for LF (0.005–0.1 Hz) and HF (0.3 – 3 Hz) ranges for the group of healthy subjects and patients at wavelengths of 687 nm (a and b) and 830 nm (c and d). The boxes indicate the 25th and 75th percentiles of the distributions. The horizontal red lines inside the rectangles denote median values and the vertical bars reflect the maximum–minimum ranges.

Fig. 6

AAF and VAF levels calculated for healthy subjects and patients with brain disorders at wavelengths of 687 nm (a) and 830 nm (b).