Header

UZH-Logo

Maintenance Infos

Shedding light on human brain activity - Biomedical signal processing of changes in tissue oxygenation and hemodynamics measured non-invasively using functional near-infrared spectroscopy: new methods and applications


Scholkmann, Felix. Shedding light on human brain activity - Biomedical signal processing of changes in tissue oxygenation and hemodynamics measured non-invasively using functional near-infrared spectroscopy: new methods and applications. 2014, University of Zurich, Faculty of Science.

Abstract

Light can be used to measure the activity of the human brain non-invasively. This is realized by shining near-infrared light into brain tissue, measuring the diffuse reflected light at different wavelengths, and determining the concentration changes of oxy- and deoxyhemoglobin ([O2Hb], [HHb]) which are related to changes in tissue hemodynamics and oxygenation – and thus to brain activity. This method, termed ‘functional near-infrared spectroscopy’ (fNIRS), is increasingly employed for basic brain research, and its routine usage for medical applications in a clinical setting is imminent.
The thesis set out to address three aims: develop and apply new approaches in order to (i) improve the fNIRS signal quality; (ii) realize advanced multivariate signal-analysis in the time-frequency domain using fNIRS signals; and (iii) investigate the systemic confounders in fNIRS studies by studying the effect of changes in partial pressure of arterial CO2 (PaCO2) on fNIRS-derived changes in brain hemodynamics and oxygenation.
The first aim was tackled by developing two novel signal processing methods that detect and remove movement artifacts (MAs) from fNIRS signals, either by using only the signal characteristic of the fNIRS input signal by itself (the ‘movement artifact removal algorithm’, MARA) or by adding signals from an accelerometer (the ‘acceleration-based movement artifact reduction algorithm’, AMARA).
Both algorithms were successfully validated. Another study investigated how different methods for determining [O2Hb] and [HHb] are affected by MAs. A systematic analysis was performed showing that multi-distance based fNIRS methods are superior to single-distance ones with regard to their robustness to MAs. In another work, a general equation was derived (based on measured data) for modeling the light-path length in human brain tissue (i.e. the differential pathlength factor, DPF) depending on wavelength and age of the subject. The equation can be used in all fNIRS applications where the light transport through tissue is modeled based on the modified Beer-Lambert law. This will improve the signal quality by minimizing the crosstalk in the determination of [O2Hb] and [HHb]. In two subsequent projects, it was shown how to extract the blood volume pulse (BVP) from fNIRS signals; and a novel approach was developed (the ‘automatic multiscale-based peak detection’, AMPD) for a precise and reliable detection of the BVP peaks.
The second aim was addressed using frameworks based on the continuous wavelet transform and Stockwell transform to quantify the relationship between fNIRS signals measured simultaneously on two human brains (the ‘hyperscanning’ approach), and between fNIRS and skin conductance signals. The value of both methods in quantifying signal correlations in the time-frequency domain was successfully demonstrated.
Finally, the third aim was realized by performing and analyzing measurements of changes in cerebral hemodynamics/oxygenation and PaCO2 dynamics in parallel during different types of speech tasks. It was demonstrated that even small changes in PaCO2 during periods of task-evoked brain activity are reflected in characteristic
changes in the fNIRS-derived signals – an observation that is of relevance for all future fNIRS studies involving speech tasks.
In conclusion, within the scope of the thesis, new fNIRS signal processing methods were developed and applied, enabling new insights into the physiological proii cesses influencing fNIRS-derived signals. The work presented in this thesis was published in ten peer-reviewed journal papers. Three additional papers have been submitted and are under review as of the submission date of this thesis. If the results presented in this thesis stimulate further research in this area and help other researchers in performing future fNIRS studies, this thesis has fulfilled its purpose.
I am convinced that spectroscopic methods with light will play an important role in future brain research and medical applications.

Abstract

Light can be used to measure the activity of the human brain non-invasively. This is realized by shining near-infrared light into brain tissue, measuring the diffuse reflected light at different wavelengths, and determining the concentration changes of oxy- and deoxyhemoglobin ([O2Hb], [HHb]) which are related to changes in tissue hemodynamics and oxygenation – and thus to brain activity. This method, termed ‘functional near-infrared spectroscopy’ (fNIRS), is increasingly employed for basic brain research, and its routine usage for medical applications in a clinical setting is imminent.
The thesis set out to address three aims: develop and apply new approaches in order to (i) improve the fNIRS signal quality; (ii) realize advanced multivariate signal-analysis in the time-frequency domain using fNIRS signals; and (iii) investigate the systemic confounders in fNIRS studies by studying the effect of changes in partial pressure of arterial CO2 (PaCO2) on fNIRS-derived changes in brain hemodynamics and oxygenation.
The first aim was tackled by developing two novel signal processing methods that detect and remove movement artifacts (MAs) from fNIRS signals, either by using only the signal characteristic of the fNIRS input signal by itself (the ‘movement artifact removal algorithm’, MARA) or by adding signals from an accelerometer (the ‘acceleration-based movement artifact reduction algorithm’, AMARA).
Both algorithms were successfully validated. Another study investigated how different methods for determining [O2Hb] and [HHb] are affected by MAs. A systematic analysis was performed showing that multi-distance based fNIRS methods are superior to single-distance ones with regard to their robustness to MAs. In another work, a general equation was derived (based on measured data) for modeling the light-path length in human brain tissue (i.e. the differential pathlength factor, DPF) depending on wavelength and age of the subject. The equation can be used in all fNIRS applications where the light transport through tissue is modeled based on the modified Beer-Lambert law. This will improve the signal quality by minimizing the crosstalk in the determination of [O2Hb] and [HHb]. In two subsequent projects, it was shown how to extract the blood volume pulse (BVP) from fNIRS signals; and a novel approach was developed (the ‘automatic multiscale-based peak detection’, AMPD) for a precise and reliable detection of the BVP peaks.
The second aim was addressed using frameworks based on the continuous wavelet transform and Stockwell transform to quantify the relationship between fNIRS signals measured simultaneously on two human brains (the ‘hyperscanning’ approach), and between fNIRS and skin conductance signals. The value of both methods in quantifying signal correlations in the time-frequency domain was successfully demonstrated.
Finally, the third aim was realized by performing and analyzing measurements of changes in cerebral hemodynamics/oxygenation and PaCO2 dynamics in parallel during different types of speech tasks. It was demonstrated that even small changes in PaCO2 during periods of task-evoked brain activity are reflected in characteristic
changes in the fNIRS-derived signals – an observation that is of relevance for all future fNIRS studies involving speech tasks.
In conclusion, within the scope of the thesis, new fNIRS signal processing methods were developed and applied, enabling new insights into the physiological proii cesses influencing fNIRS-derived signals. The work presented in this thesis was published in ten peer-reviewed journal papers. Three additional papers have been submitted and are under review as of the submission date of this thesis. If the results presented in this thesis stimulate further research in this area and help other researchers in performing future fNIRS studies, this thesis has fulfilled its purpose.
I am convinced that spectroscopic methods with light will play an important role in future brain research and medical applications.

Statistics

Downloads

112 downloads since deposited on 15 Jan 2015
35 downloads since 12 months
Detailed statistics

Additional indexing

Item Type:Dissertation
Referees:Pfeifer Rolf, Wolf Martin
Communities & Collections:04 Faculty of Medicine > University Hospital Zurich > Clinic for Neonatology
Dewey Decimal Classification:610 Medicine & health
Language:English
Date:2014
Deposited On:15 Jan 2015 08:35
Last Modified:27 Apr 2017 22:48
Number of Pages:329

Download

Download PDF  'Shedding light on human brain activity - Biomedical signal processing of changes in tissue oxygenation and hemodynamics measured non-invasively using functional near-infrared spectroscopy: new methods and applications'.
Preview
Content: Other
Language: English
Filetype: PDF (Abstract/Zusammenfassung der Dissertation)
Size: 2MB