Task 7 Flashcards
Neurovascular coupling (first method)
Neurovascular coupling = A reduction in local glucose and oxygen stimulates the brain to increase local arteriolar vasodilation, which increases local cerebral blood flow (CBF) and cerebral blood volume (CBV)
Glucose and oxygen are carried to the active area in need
Because oxygenated (oxy-Hb) and deoxygenated hemoglobin (deoxy-Hb) have characteristic optical properties in the visible and near-infrared light range (700 – 900), the change in concentration of these molecules during neurovascular coupling can be measured using optical methods.
2 physiological events caused by brain activity can be assessed using optical techniques.
Neurovascular coupling = A reduction in local glucose and oxygen stimulates the brain to increase local arteriolar vasodilation, which increases local cerebral blood flow (CBF) and cerebral blood volume (CBV) Glucose and oxygen are carried to the active area in need Because oxygenated (oxy-Hb) and deoxygenated hemoglobin (deoxy-Hb) have characteristic optical properties in the visible and near-infrared light range (700 – 900), the change in concentration of these molecules during neurovascular coupling can be measured using optical methods. Most common measurement in fNIRS: measurement of changes in the ratio of oxy-Hb to deoxy-Hb The wavelengths reflected by oxy-Hb and deoxy-HB are to be found in a specific range, also called the “optical window”. Photons introduced at the scalp pass through the most of the tissue, and are either absorbed, scattered, or reflected back from oxy-Hb and deoxy-Hb. Because relatively predictable quantities of photons follow a banana-shaped path back to the surface of the skin, these can be measured using photodetectors. By measuring absorbance-reflectance changes at two (or more) wavelengths, one of which is more sensitive to oxy-Hb, the other to deoxy-Hb, changes in the relative concentration of these chromophores can be calculated
The hesitation over utilization of fNIRS may be due to the current limitations in emergent technology. These include
- limited spatial resolution
- limitations in the use of cranial reference points
- attenuation of the light signal by extracerebral matter
- comparisons of fNIRS data between subjects
- the impact of skin pigmentation on signal detection
- difficulties obtaining absolute baseline concentrations of oxy-Hb and deoxy-Hb
The fNIRS apparatus
- Light source / optode (LED or fiber-optical bundles)
- Light detector / photodetector that receives light after it has been reflected from the tissue (placed 2-7cm away from the optode) (there are less detectors than sources)
o When placed 4cm from the optode: fNIRS is sensitive to hemodynamic changes (within the top 2-3mm of the cortex)
o 2 – 2,5 distances: gray matter is part of the sample volume
o Umso weiter die distanz zwischen optode und detector, desto teifer kann mann messen aber umso ungenauer wird es auch
Three distinct types of fNIRS implementations have been
; time-resolved systems, frequency domain systems, and continuous wave spectroscopy measurements
- Time resolved & frequency domain can give information about phase and amplitude and are necessary when we want to have precise quantifications of fNIRS signals
- Continuous wave fNIRS system (CW): apply light to tissue at constant amplitude, measuring the attenuation of amplitude of the incident light. Even though they provide generally less information than the other techniques, they are useful in some cases.
o Use LEDs (safe for eyes)
o Cheap (used in clinical settings)
o Very small and portable
Event-related optical signal (EROS)
Sensitive to changes in the optical properties of the cell membranes themselves that occur as a function of the ionic fluxes during firing (ionic fluxes are usually recorded by EEG).
Measures the depolarization state of neuronal tissues (direct measure & non-invasive)
Even though this method has the big advantage of a high temporal resolution and superior spatial resolution to EEG, EROS has limitations:
- Low signal-to-noise ratio (SNR) resulting from the need to image through skin, skull, … we therefore need many trials (500-1000) to establish a reliable signal (we need a high SNR)
- When there is a visual stimulus, we cannot replicate a fast optical signal (in response to a visual stimulus)
- Requires a more expensive and cumbersome laser-based light source (versus an LED-based light source). Laser-based methods have the advantage of measuring both amplitude and phase
o Limitation: not portable & increased potential risk of damage to the eyes
fnirs quick facts
measure Changes in the ratio of oxy-Hb to deoxy-Hb (dependent on the hemodynamic response)
Indirect measurement
spatial resolution: Only 1cm2 due to the scatter of photons (similar to EEG but a bit better)/worse teh firms
Capacity is limited to outer cortex
Participants can sit upright , work on a computer
is not susceptible to movement artefact
inexpensive
easily used in children
Fnirs and firm similarities
Temporal resolution is limited by the hemodynamic response (takes a few seconds)
Safe, non-invasive, can be used repeatedly on the same individual
Require repeated stimulation due to signal-to-noise ratio
FnIRS limitations
- limited spatial resolution
- use of cranial reference points
- attenuation of the light signal by extracerebral matter
- comparisons of fNIRS data between subjects
- the impact of skin pigmentation on signal detection
- difficulties obtaining absolute baseline concentrations of oxy-Hb and deoxy-Hb
- failures to replicate
- high signal-to-noise ratios in EROS signals
- its early developmental stage
- need for informatics-driven modifications of clinical databases based on standardized and normative data
the major benefit of fNRI
its application extends beyond the confines of traditional neuroimaging laboratory settings and instead can be extended to various clinical populations to conduct ‘‘real-world’’ monitoring of cognitive function in ‘‘real time’’ for extended periods in a continuous and safe manner for purposes of both research and clinical interventions.
Examples: people suffering from schizophrenia when they experience hallucinations or delusions
fNIRS can be extended to populations that were previously understudied (i.e. due to limitations of weight, height, claustrophobia, young age, …).
Time-course related changes can also be examined with fNIRS.
Example: real time measures of oscillations between depressed and manic states
fNIRS could also be implied in treatment or substance abuse (neurofeedback, …), especially when combined with neurorehabilitation, rTMS, ECT and CBT.
direct neural interafce
= hardware and software that provide direct communication between the neural activity of the brain and computer components without the involvement of peripheral nerves and muscles
The BCI and the BMI (brain-machine interface) are direct neural interfaces.
EEG signal detection
Differences in electric potential on the scalp (the sum of excitatory and inhibitory post-synaptic potentials) are measured
MEG signal detection
SQUIDS detect magnetic disturbances in the brain. This method suffers from the lack of spatial resolution and difficult testing conditions
FMRI signal detection
Measured the BOLD signal response to acquire a series of susceptibility weighted images (covering the whole or part of the brain) during block-based or event-related behavioural or cognitive task. Ultra-high-field (UHF) scanners (>7T) are becoming more available for potential BCI applications.
NIRS signal detection
Measures the ratio of deoxy-Hb and oxy-HB in the blood through the differences in light absorption by the 2. Multiple “optodes” that emit and detect the light, are placed around the scalp
ftcd signal detection
Based on ultrasound doppler imaging (initially developed to measure the velocity of blood flow in major cerebral arteries). Ultrasonic wave frequencies that are reflected by the blood are detected. This method has limited depth penetration (due to skull) and the examination is constrained to the major vessels (i.e. detect the hemispheric dominance in blood flow during a language task)
HMI
: During the execution of a motor task, specific areas of the brain are activated, the neural signal is transmitted to the muscles. This movement is relayed to HMIs (i.e. computer mouse) and generated a specific input for the computer. This sensory information is fed back to the operator, thus forming a closed loop (helpful for motor learning etc.).
BCI
involvement of movement (motor pathways) is replaced by the BCI process, in which the neural signals are translated onto specific machine commands by a set of rule-based algorithms. These commands are often aided by machine-learning algorithms to improve the accuracy of prediction.
P300 component if often used for BCI because of its association with categorical stimulus-evaluation
processes. Another component being used are slow-cortical potentials and steady-state visual-evoked potentials.
problem of bci and meg
This application faces difficulties:
- poor spatial resolution of EEG results in an inverse problem (we must rely on computational methods to localize the source of the activation)
- a sufficient training phase is required (due to variance in spontaneous EEG activity – can be reduced through machine learning)
MEG-based BCI
MEG signals are magnetic ‘counterparts’ of EEG signals (so MEG BCIs are similar to EEG BCIs in that they employ similar data processing strategies). 69% classification accuracy for spatial attention, 95– 97% classification accuracy for motor tasks and 86–87% classification accuracy for motor imagery tasks have been observed in MEG-based BCI (no lengthy training needed as opposed to EEG).
Advantages:
- magnetic fields are less distorted than electrical fields (MEG signals are less affected by properties of the skull)
o better spatial resolution & and SNR
- provides consistent feedback experience and faster learning of rhythm control for participants (due to better signal-to-noise ratio than EEG).
fMRI-based BCI
because fMRI can be obtained in real-time (rtfMRI), it is a valuable technique for BCI.
Example: 2D movement of a robotic arm was controlled by the regulation (and concurrent detection) of regional cortical activations in the primary motor areas. In this study, the participants engaged in right- and/or left-hand motor imagery tasks, and the BOLD signals originating from the corresponding hand motor areas were then translated into horizontal or vertical robotic arm movement
Automated interpretation of fMRI data (automatic pattern recognition) is currently the main goal for fMRI-BCI and are being tested and developed through machine-learning paradigms.
advantage: non-invasive, high spatial resolution
limitation: high cost, patient musst lay still, susceptible to movement artefacts
NIRS-based BCI
Using motor imagery with a NIRS system, 89% correct classification of right and left hand imagery tasks has been observed.
NIRS can also detect the P300 component and provide accurate classifications and predictions (about 80%).
Advantages:
- flexibility of use
-less sensitive to movement artefacts
-no restriction on paradigm,
- portability
- metabolic specificity
- high sensitivity in detecting small substance concentrations
- affordability
- requires neither conductive gel nor corrosive electrodes, making it suitable for extended use
The major disadvantage of NIRS is the latency of the hemodynamic response, resulting in slow operation of the NIRS-BCI system as well as the inability to characterize the signals from subcortical regions. To improve the feasibility of NIRS-based BCIs, the influence of respiration and blood pressure on hemodynamic response has to be reduced and higher spatial resolution needs to be attained.
EEg & BCI
advantages: noninvasive, portable, silent
disadvantages: limited spatial resolution, susceptible to artifacts from cranial muscles and eye movement
SORGER COMMENTARY ON NAGELS-COUNE – BRAIN-BASED COMMUNICATION VIA ONLINE-DECODED FNIRS SIGNALS
An easy yes/no communication paradigm relying on mental imagery and fNIRS was implemented: mental drawings (imagine drawing simple geometric figures) encoded “yes” and not changing mental state encoded the answer “no”#
. This method might therefore also be useful for non-responsive patient
preprocessing methods
serve to get rid of frequencies of no interest in the measured signal (i.e., to clean up the data). The frequency of interest here was the rate at which the participant responded, which was five times per run.
2 types of analysis were conducted: univariate general linear model (GLM) and multi-channel pattern (MCP).
- GLM:
o single channel of interest (best channel) was determined by calculating a criterion value by averaging the obtained oxy-Hb and deoxy-Hb t-values per channel (The channel yielding the largest t-value (averaged across oxy-Hb and deoxy-Hb) was defined as the ‘best’ channel).
o Data of first YES / NO run was analysed by calculating the individual criterion value for each trial. A mean was computed and used as cut-off value for the remaining runs.
o Encoded answers were compared post hoc to the actually intended answers given by the participant.
So the first answer-encoding run was ‘sacrificed’ to identify the fNIRS response amplitude during “yes” and during “no—the midpoint between these two amplitudes could then be used as a classification criterion in the other runs: fNIRS response amplitudes falling above (below) this criterion were classified as a “yes” (“no”) response, respectively.
results
A best channel was obtained for each subject. With the GLM, answers could be decoded correctly with an average accuracy of 64.25% (which is higher than the chance level of 50%)
Comfortability rating of the subjects decreased across time and dropped considerably for the last run. This was probably due to participant getting bored or tired.
discussion
The current approach has the potential of enabling immediate communication in the order of ca. 30 min. This specific design could be difficult to perform in non-responsive patients, because of the importance of the localizer run (the run that contains differential activity (mental imagery vs. rest) to train the classifier).
Some participants performed exceptionally well whereas for others classification accuracy was at chance level, so the effectiveness of the BCI depends on the participant, which is a potential obstacle for application in patients: it cannot be easily distinguished whether a patient yielding a poor decoding accuracy is unconscious or simply a rather ineffective encoder. More research is required to identify the cause of this inter-individual variability
fnri
Changes in electrochemical activity and the consequent change in blood levels affect its optical properties
Can measure changes in ratio of oxy-Hb & deoxy-Hb
Most biological tissues are relatively transparent to light in near-infrared range (spectrum/wavelengths: 700-1000nm)
BUT oxy-Hb & deoxy-Hb have diff optical properties diff NIR light absorption spectra each reflects different wavelength within near-infrared range
This spectral band is often referred to as optical window
Oxy-Hb = 760, deoxy-Hb = 850
better ability to filter out noise
how does it work
- Place sources of light & detectors on head surface
Optode – light source (e.g. LEDs)
Light detector – receives reflected light (duh…) - When we introduce photons at the scalp, they pass through most of the tissue but some are reflected by oxy-HB & deoxy-HB
80% of photons follow banana-shaped path from source to detector
Others get absorbed or are scattered but don’t reach detector (see pic) - Changes in the Hb-concentrations causes changes in the reflected light intensities
This is quantified using Beer-Lambert law empirical description of optical attenuation in a highly scattered medium (cool, but this info gives me nothing…) - fNIRS measures alterations in the intensity of attenuated light at different wavelengths
Measure absorbance/reflectance changes at the 2 different wavelengths changes in the relative concentration of Hbs can then be calculated
Depth estimated by time-of-flight distributions
fnris
Functional near infrared spectroscopy (fNIRS) has the attributes necessary to translate laboratory work into more realistic everyday settings and clinical environments. It offers relatively non-invasive, safe, portable low-cost methods of both direct and indirect monitoring of brain activity.
Because it is compact and portable, it allows us to investigate clinical and social environments and it can readily be integrated into other technologies like EEG or TMS.
However, we are still quite unfamiliar with fNIRS and it does have spatial limitations relative to fMRI.
fmri
measure indirectly BOLD response
better spatial resolution than Fnris
can imagine the entire
brain
particpant must lie inside magnet that is very loud
very sensitive to movements
expensive
Because fNIRS is such a new technique, we must establish
Because fNIRS is such a new technique, we must establish it with regard to other functional neuroimaging techniques like fMRI (sensitivity, accuracy, reliability, …):
- High correlations between BOLD and deoxy-Hb concentration: strong correlations between fMRI changes and all optical measures, with oxy-Hb/deoxy-Hb providing the strongest correlation.
- In a group of participants, good reliability and excellent reproducibility of the fNIRS were reported
- Obrig et al. (2000): demonstrated concordance between fNIRS and fMRI signals using visual stimulation and a response inhibition task
preliminary work in our laboratory suggests that there is some congruency in the signal for fNIRS and fMRI.
data analysis -sorger
- Preprocessing: the measured light intensities were transformed into oxygenated hemoglobin (oxy-Hb) and deoxygenated hemoglobin (deoxy-Hb) values. Linear trend removal, temporal lowpass filtering and high-pass filtering (cutoff: 10 cycles [localizer runs] or 2 cycles [answer encoding runs] per time course) were applied.
- Subsequent data analysis: focused on 14 neighbour channels (i.e., channels emerging from sources-detector combinations of close proximity). These channels were used only for r analysis of the localizer data and when following the multivariate (MCP analysis) approach. In the univariate analysis of the main answer-decoding data it was focused on a single (the ‘best’) channel, obtained through preprocessing.
