A review on continuous wave functional near-infrared spectroscopy and imaging instrumentation and methodology

Highlights

• Comprehensive review on continuous wave functional near infrared imaging

• Overview of currently available commercial near infrared imaging instrumentation

• Review of technical aspects such as light sources, detectors and sensor arrangements

• Review of methodological aspects, algorithms, and data analysis and its tool boxes

Abstract

This year marks the 20th anniversary of functional near-infrared spectroscopy and imaging (fNIRS/fNIRI). As the vast majority of commercial instruments developed until now are based on continuous wave technology, the aim of this publication is to review the current state of instrumentation and methodology of continuous wave fNIRI. For this purpose we provide an overview of the commercially available instruments and address instrumental aspects such as light sources, detectors and sensor arrangements. Methodological aspects, algorithms to calculate the concentrations of oxy- and deoxyhemoglobin and approaches for data analysis are also reviewed.

From the single-location measurements of the early years, instrumentation has progressed to imaging initially in two dimensions (topography) and then three (tomography). The methods of analysis have also changed tremendously, from the simple modified Beer-Lambert law to sophisticated image reconstruction and data analysis methods used today. Due to these advances, fNIRI has become a modality that is widely used in neuroscience research and several manufacturers provide commercial instrumentation. It seems likely that fNIRI will become a clinical tool in the foreseeable future, which will enable diagnosis in single subjects.

Introduction

Continuous light has been used to non-invasively investigate human tissue such as the breast, head and testes by transmitting the light through the body as early as at least in the 19th century (Bright, 1831, Curling, 1856, Cutler, 1929). More specifically, already in 1862 Hoppe-Seyler from Germany, described the spectrum of oxy-hemoglobin (O₂Hb) (Perutz, 1995) and in 1864 Stokes from the United Kingdom added the spectrum of deoxy-hemoglobin (HHb) and consequently discovered the importance of hemoglobin for the oxygen transport (Perutz, 1995). In 1876 von Vierordt, also from Germany, analyzed tissue by measuring the spectral changes of light penetrating tissue when the blood circulation was occluded (Severinghaus, 2007, von Vierordt, 1876) and in 1894 Hüfner from Germany spectroscopically determined absolute and relative amounts of O₂Hb and HHb in vitro (Hüfner, 1894). After decades of no relevant research in this field, in the 1930s the work on spectroscopic determination of tissue oxygenation was continued by several researchers. For example Nicolai, Germany, repeated the study of von Vierordt (Nicolai, 1932a, Nicolai, 1932b), and Matthes and Gross, Germany, demonstrated for the first time the spectroscopic determination of O₂Hb and HHb in human tissue using two wavelengths, one in the red and near-infrared region (Matthes and Gross, 1938a, Matthes and Gross, 1938b, Matthes and Gross, 1938c).

In terms of quantification, an important first step was the discovery of the Beer–Lambert law first by the French mathematician Bouguer in 1729 (Bouguer, 1729). It is often attributed to the Swiss Lambert, although he cited Bouguers work in 1760 himself (Lambert, 1760). The law was extended by the German Beer to quantify concentrations in 1852 (Beer, 1852). Since the Beer–Lambert law is only valid in non-scattering media, it cannot be applied to biological tissue. Relatively recently therefore the modified Beer–Lambert law (MBLL) was developed by the British Delpy (Delpy et al., 1988), to take into account the light scattering. The MBLL is often used by many instruments described in this review. Further important steps were also analytical solutions of the diffusion equation (e.g. Arridge et al., 1992, Patterson et al., 1989) to quantitatively describe light transport in tissue.

Based on the insight of the relative transparency of the tissue including the skull in the near infrared range in 1977 Jobsis from the USA first demonstrated the feasibility to continuously and non-invasively monitor the concentration of O₂Hb and HHb ([O₂Hb] and [HHb]) in the brain (Jobsis, 1977). Therefore he is considered to be the initiator of near-infrared spectroscopy (NIRS).

His discovery led to designing and building of several NIRS instruments (Ferrari and Quaresima, 2012). All these instruments were continuous wave (CW) instruments. The term “continuous wave” means that the instrumentation is solely based on a light intensity measurement, i.e. near-infrared light is sent into the tissue and the intensity of the re-emerging (i.e. diffusely reflected) light is measured. This is in contrast to time resolved techniques such as time and frequency domain techniques, which, additionally to the intensity measurements also measure the time of flight, i.e. the time that the light needs to travel through the tissue. For a visualization of the three different techniques please refer to Fig. 1.

The disadvantage of CW systems is that they cannot fully determine the optical properties of tissue (i.e. light scattering (μ_s′) and absorption (μ_a) coefficients) and therefore the [O₂Hb] and [HHb] cannot be determined absolutely. However, with a few reasonable assumptions it is possible to quantify changes in [O₂Hb] and [HHb]. Therefore, during the first years, NIRS instruments were mostly trend monitors, employed to study various physiological conditions and clinical interventions. Much research was aimed at obtaining absolute values either by physiological maneuvers (e.g. Edwards et al., 1988, Wyatt et al., 1990) or enhancing the instrumentation (e.g. Matcher et al., 1995a, Matcher et al., 1995b, Wolf et al., 1997). Later time resolved techniques were developed and became available and enabled to determine absolute values. This will not be discussed further, because it is not within the scope of this review.

1993 was a crucial year in the development of functional NIRS (fNIRS) of the brain. In the same year four research groups published results and demonstrated that it is possible to non-invasively investigate brain activity using fNIRS (Chance et al., 1993, Hoshi and Tamura, 1993, Kato et al., 1993, Villringer et al., 1993). Brain activity leads to an increase in oxygen consumption, which is accompanied by an increase in cerebral blood flow due to neurovascular coupling. This leads to a change in the local [O₂Hb] and [HHb] (Wolf et al., 2002), which can be detected non-invasively by fNIRS. These first measurements were carried out with simple instruments, which measured at one or a few locations. Since brain activity in response to a stimulation occurs only at specific locations in the brain, when measuring just at one location it is often difficult to find the correct position on the head for the measurement. In addition, there is a scientific interest in measuring a spatial pattern, how the brain activity affects an area of the brain.

A next major step in the development was to design imaging instruments that covered a larger area of the head and enabled mapping of brain activity, i.e. to deliver topographic information (Ferrari and Quaresima, 2012, Maki et al., 1995). This had several advantages: it enabled to localize brain activity and the precise localization of the sensor was less important. This technology is called functional near infrared imaging (fNIRI). On the one hand, it was quite clear that it is highly important to expand the interrogated area by using imaging systems. On the other hand, quantification is not that important in neuroscience, i.e. it is more important to statistically significantly detect a change in brain activity than to quantify it in absolute terms. For these reasons up to today most imaging systems are based on CW technology. In addition, time resolved systems have a lower time resolution, are more expensive and the time of flight is generally a more noisy parameter than the intensity and therefore not useful for detecting small functional activations. In contrast, CW systems are relatively low cost, can be miniaturized and wireless systems and can be applied unobtrusively in everyday life situations or even freely moving animals (Muehlemann et al., 2008).

As a next step, sensor arrangements where several source detector distances are measured simultaneously, so-called overlapping measurements, enabled to apply tomographic approaches, i.e. image reconstructions in three dimensions (Joseph et al., 2006).

Today fNIRI has entered neuroscience as a research tool. It has been shown that fNIRI is reliable and trustworthy for research based on investigating groups of subjects, although reliability in single subjects is not sufficient yet (Kono et al., 2007, Plichta et al., 2006, Plichta et al., 2007a, Plichta et al., 2007b, Schecklmann et al., 2008). Consequently, the number of publications on fNIRI in neuroscience has increased exponentially within the last years.

One of the next aims is to apply fNIRI clinically. For this purpose it will be compulsory to ensure a high reproducibility in single subjects. However, a sufficient reliability on the single subject level has not yet been achieved (Biallas et al., 2012a, Biallas et al., 2012b, Kono et al., 2007, Plichta et al., 2006, Plichta et al., 2007a, Plichta et al., 2007b, Schecklmann et al., 2008). Hence, some research is currently focused on improving the reliability. Possible reasons for the lack in reliability are the superficial tissue (i.e. light has to penetrate several tissue layers such as e.g. skin and skull before it reaches the brain) or systemic physiological changes, which contaminate the signal of the brain and possibly instrumental shortcomings such as an insufficiency in spatial resolution and/or signal to noise ratio (SNR). Generally there are several possibilities to improve reliability: On the instrumental level, it is important to select appropriate wavelengths, light sources, detectors, and geometrical arrangements to avoid crosstalk and ensure a high SNR; on a methodological level the aim is to reduce the influence of superficial tissue or systemic components.

The aim of this paper is to review the current state of instrumentation and methodology of CW fNIRI. For this purpose we will give an overview of the commercially available instruments and address instrumental aspects such as light sources, detectors and sensor arrangements. Methodological aspects such as algorithms to determine [O₂Hb] and [HHb] and data analysis will also be reviewed.

There is a variety of terms used for NIRS. In general the term NIRS is often used as an overarching term for the whole technology, but in principle it only refers to NIRS systems measuring at single locations with up to four sensors, but without imaging capacity. Imaging systems, which we will call NIRI here, have more than 4 channels and produce two or three dimensional images. In the literature, 2D imaging systems are also called near infrared topography or mapping, or diffuse optical imaging. 3D imaging systems are also called near infrared tomography or mapping, or diffuse optical tomography or imaging. The term “diffuse” refers to the fact that due to the high μ_s′ in tissue the propagation of photons through tissue can be modeled as a diffusion process. Unfortunately, the term “diffuse” is often misinterpreted as blurred or fuzzy, i.e. referring to a negative connotation of the technique, which clearly is not meant or adequate. As for magnetic resonance imaging (MRI), which was formerly referred to as nuclear magnetic resonance, the term “nuclear” falsely evoked the association of a radioactive method, we propose to abandon the term “diffuse” in the future to avoid this negative connotation. The spatially resolved measurement of brain activity in two and three dimensions we call functional NIRI (fNIRI). All systems (except for one) and methods reviewed here are generally about CW fNIRI and therefore the term “CW” will be omitted below.