Recovering fNIRS brain signals: physiological interference
suppression with independent component analysis
Y. Zhang
∗,a
, M. Shi
b
, J. Sun
a
, C. Yang
a
, Y.J. Zhang
c
, F. Scopesi
e
, P. Makobore
f
, C. Chin
g
, G. Serra
e
,
Y.A.B.D. Wickramasinghe
d
, P. Rolfe
a,d,e
a
School of Electrical Engineering and Automation, Harbin Institute of Technology, Harbin, China
b
The First Affiliated Hospital of Harbin Medical University, Harbin, China
c
Harbin Branch of Hoau Cooperation, Harbin, China
d
Oxford BioHorizons Ltd, Maidstone, United Kingdom.
e
Neonatal Intensive Care Unit, Istituto Gianina Gaslini , University of Genova, Genova, Italy.
f
Uganda Industrial Research Institute, Kampala, Uganda
g
Functional NeuroRehab Centre Pte Ltd., 42 Jalan Pemimpin, Singapore
ABSTRACT
Brain activity can be monitored non-invasively by functional near-infrared spectroscopy (fNIRS), which has several
advantages in comparison with other methods, such as flexibility, portability, low cost and fewer physical restrictions.
However, in practice fNIRS measurements are often contaminated by physiological interference arising from cardiac
contraction, breathing and blood pressure fluctuations, thereby severely limiting the utility of the method. Hence, further
improvement is necessary to reduce or eliminate such interference in order that the evoked brain activity information can
be extracted reliably from fNIRS data. In the present paper, the multi-distance fNIRS probe configuration has been
adopted. The short-distance fNIRS measurement is treated as the virtual channel and the long-distance fNIRS
measurement is treated as the measurement channel. Independent component analysis (ICA) is employed for the fNIRS
recordings to separate the brain signals and the interference. Least-absolute deviation (LAD) estimator is employed to
recover the brain activity signals. We also utilized Monte Carlo simulations based on a five-layer model of the adult
human head to evaluate our methodology. The results demonstrate that the ICA algorithm has the potential to separate
physiological interference in fNIRS data and the LAD estimator could be a useful criterion to recover the brain activity
signals.
Keywords: Functional near-infrared spectroscopy, physiological interference, modified lambert-beer law, independent
component analysis, least-absolute deviation
1. INTRODUCTION
Functional near-infrared spectroscopy (fNIRS) is a noninvasive technique for assessment of oxygenation of the human
brain both in adults and infants using light in the 650 to 950-nm wavelength range
[1]
. With continuous-wave systems,
changes in oxyhaemoglobin and deoxyhaemoglobin (HbO
2
and HHb) can be obtained using the modified Lambert-Beer
law based on changes in light absorption at multiple wavelengths
[2]
. fNIRS is an effective technique for measuring brain
activation. It has several advantages, such as flexibility, portability, low cost and fewer physical restrictions compared
with other imaging techniques, such as functional magnetic resonance imaging (fMRI), magneto encephalography
(MEG), positron emission tomography (PET), and electroencephalography (EEG)
[3-4]
. However, there is often
undesirable interference observed in fNIRS measurement data. The well-known physiological interference, which is
often relatively large and has low correlation with the functional response evoked by the task execution, is caused by
cardiac events, breathing and blood pressure pulsations. Thus, physiological interference suppression is significant in
order that the real haemodynamic response can be revealed.
In brain activation studies with fNIRS, the useful signal is in the deeper regions of the brain and the strong mixture
between the physiological interference and the brain activity response presents significant challenges in signal extraction.
∗
corresponding authors, email: zyhit@hit.edu.cn; peterrolfe@aol.com.
Ninth International Symposium on Precision Engineering Measurement and Instrumentation
Proc. of SPIE Vol. 9446 944604-1