fNIRS - Science method

Hi,

To calculate dataset features for fNIRS focused on the SMA and MI areas, first preprocess the data to remove noise and artefacts, bearing in mind that experiment design, data quality, and research questions can influence these preprocessing and analysis steps. Segment the cleaned data to isolate motor execution and imagery tasks. Optionally apply filters for further refinement. Then, utilise the fNIRS-SPM toolbox for statistical analysis and feature extraction, such as mean, variance, and functional connectivity metrics. Proper probe placement, signal quality, and correct statistical thresholds are crucial for accurate results.

Hope this helps.

fNIRS can be used to measure changes in cerebral blood flow by following method:-

Maybe this paper will help here: https://pubmed.ncbi.nlm.nih.gov/17258472/

cheers

Michael

Check with the Georgia Tech Behavioral Medicine department. I know they were doing studies on individuals' gaits.

Dear Zilu Liang,

I think you will find your answer in this blog post.

Kind regards

Hey guys,

check our research for this topic:

yes, that's why I thought MEG/EEG would be a good method. They can still speak while they can not speak in the MRI scanner.

Hi, interesting question. Could you provide more information for the search of the mentioned articles (ciftni et al 2008 and kamran et al 2015 , kamran et al 2016 )?,

Thanks

This is a known issue when dealing with multi-channel or multi-voxel brain imaging. The main problem arises from the statistical dependance of the results as a function of space, namely, close channels or voxels are highly correlated. This can drastically inflate your false negatives when using standard procedures for multiple comparison correction. In my opinion the best approach is to actually account for spatial statistical dependence using, for example, gaussian random field theory. The approach will provide an estimation of the 'effective' number of independent measures. This is the standard procedure in fMRI where there are a huge amounts of voxels to deal with. If i remember correctly, nirs_spm had an implementation of this approach, however the software is getting outdated.

Michael M Plichta thank you very much~

Dear Mehr Qorbani,

I think that the idea of ​​using a cycle ergometer is good and can be combined with BDNF. But there is one important condition related to neuroplasticity - its goal is to improve not only the speed of movements, but most likely the implementation of complex (complex) movements - accuracy, coordination, smoothness and switchability of executions ... In this regard, you must evaluate and execute real complex movements in terms of accuracy, volume, coordination. It could be hitting the goal, jumping, grabbing balls, performing alternate, equal sports movements, etc. You know better what you can turn on.

The tools you use really depend on your research question and what aspects of neuroplasticity you aim to explore. Are you looking for specific localization of brain regions, or changes in the timing of neural activity? Do you plan on having your participants complete motor activities during your measurements? Your research question(s) should guide your study design and, therefore, the tools that you use. Can you provide more detail about what you hope to study?

Dear Iris,

I just wanted to add to the interesting answer by Mathijs that, in most of fNIRS recording systems, raw absolute values are computed with reference to a what is recorded during a calibration phase.

Therefore if relevant factors (such as environmental light, posture - i.e. sitting vs. standing, ...) differ between calibration and baseline-/task-related recordings, raw absolute values could even be negative.

Because of that, in fNIRS investigations (as in other functional imaging ones) it is preferable to work with O2Hb/HHb/totHb change scores weighted on recording-specific baselines.

Hope this helps.

Best,

Davide

For further reading, I would also suggest:

- Balconi, M., & Molteni, E. (2016). Past and future of near-infrared spectroscopy in studies of emotion and social neuroscience. Journal of Cognitive Psychology, 28(2), 129–146. https://doi.org/10.1080/20445911.2015.1102919

- Crivelli, D., & Balconi, M. (2017). Near-infrared spectroscopy applied to complex systems and human hyperscanning networking. Applied Sciences, 7(9), 922. https://doi.org/10.3390/app7090922

- Ferrari, M., & Quaresima, V. (2012). A brief review on the history of human functional near-infrared spectroscopy (fNIRS) development and fields of application. NeuroImage, 63(2), 921–935. https://doi.org/10.1016/j.neuroimage.2012.03.049

It depend... If you need a portable and wireless device, I would say ARTINIS. If you need a "classical" NIRS, ARTINIS and HITACHI are probably the best options (from my point of view of course!)

Best,

This subject had no hemoglobin changes when solving only the problem(cognitive task)

But, hemoglobin changes occurred when slept or stop breath.

My device is fixed gain and light intensity enough to observe cortex.

Bring your reference electrode closer to the recording electrode.  Don't put it in the neck, put it on the skull, that removes most movement, heart beat, breathing etc artifacts. Inn mice in particular, this artifact shows up.  In rats, putting the reference in the neck is not a problem.

You can compare between the two different states and make the analysis as a % of the original signal. While MVC is the more common approach there are several studies that have used this procedure and it is also discussed in Essentials of Electromyography. Kamen & Gabriel, 2010 (Human Kinetics is the Publisher) ISBN: 9780936067126

Hi Agbangla, Not sure if I understand your question... fNIRS does not influence the delay (or cognitive function). NIRS does record the delay (it is physiology). If you would expect any differences in delay between groups, then this will be possible to measure with NIRS.

In this paper they explain how to thin the skull.

1. Neuroimage. 2016 Jan 1;124(Pt A):752-61. doi: 10.1016/j.neuroimage.2015.09.037. Epub 2015 Sep 28.

Transcranial functional ultrasound imaging of the brain using microbubble-enhanced ultrasensitive Doppler.

Errico C1, Osmanski BF1, Pezet S2, Couture O1, Lenkei Z2, Tanter M3

Mehmet

I am sorry but you are wrong - growth modeling can be performed in a multilevel framework  with individuals at level 2 and repeated measures at level 1. In fact the multilevel approach is more flexible than a structural equation approach (such as Amos) as you can include continuous time in the model

This paper does the comparison

Steele F. (2008) Multilevel Models for Longitudinal Data. Journal of the Royal Statistical Society, Series A, 171, 5-19.

Indeed the interface in SPSS for multilevel models is set up for the analysis of longitudinal data as growth models.

Dear Sir

Andrew Benjamin Gundra,

I am working on fNIR for almost 2 years. Once upon a time I also seek this question. As far I learnt from my research and the other literature based on fNIR, everyone uses oxy, de-oxy, or total oxygenation change or both oxy and deoxy data. But in all case, they used the data that gave them significant result or significant difference. By ANOVA or U test you may not get any significant differences in case of oxy data, then you must try for deoxy or total change in oxygenation or deoxygenation. Some times both results are combined to make the information meaningful. I think this is just to convince and make the research meaningful.

If I become wrong in my opinion, I beg your pardon in advance.

Thanks

Md. Asadur Rahman

BME, KUET, Khulna, Bangladesh.

Our software package, EventIDE (www.okazolab.com) can alone handle all functions that you need. That would include stimulus presentation, eye-tracking recording (with any eye-tracker mode) and synchronization with fNIRS (up to millisecond precision).  In addition, you can monitor and process eye-tracking data and brain signals in real-time, for example, allowing stimulus adjustment based on instant brain-eye coherence, as Adam suggested.

We make sure that we see the heart beat in the signal and that the heart beat is stronger in the oxygenated hemoglobin signal than the deoxygenated hemoglobin signal. The better the contrast to noise ration for the heart beat the better off you are. I like it when we have a cnr of 10 or better. Likewise, the modulation in oxy should be 5x or more larger than deoxy 

I think this is not so easy to consider the baseline for all the modalities. Therefore, you have to consider a general hypothesis to measure the signals because their temporal resolutions are not same. Although the baseline of these three signals are not same.

Dear Mam,

As far I know, the series of NIRS-SPM give their specification to convert the fNIR data to their format. I think you should read the manual carefully. If the problem arise again, I can try to help you.

You cannot compute O2 saturation that way. From what I gathered, you are using the modified Lambert-Beer law to compute 'relative' changes in concentrations from the changes in light intensity. As the previous sentence implies, and Michael Nordine already correctly pointed out, These changes are always *relative* with respect to some moment in time. This means, you can add an arbitrary number to it, and still get the a valid 'relative' change (e.g. an increase of 1 a.u. in oxygenated hemoglobin, no matter if your values are 2 and 1, or 102 and 101). Computing the ratio between HbO and HbT (which is HbO+HbR) is dependent on the absolute values (because 1/2 is different than having added 100 to both, i.e. 101/102). This 'arbitrariness' is an inherit characteristic of the modified Lambert-Beer law and thus cannot be circumvented. Hence, there is no way to get the absolute oxygen saturation for these measurements (assuming this is what you did). For example, you can also get different HbX values if you pump up your laser power, resulting in more light reaching the receiver. In absolute terms, however, there should have been no change.

In order to measure absolute oxygenation, you need a different technique. We use the tissue saturation index (TSI), which for example our PortaLite devices can measure, but also our versatile, multichannel Oxymon devices are able to do. See e.g.

Feel free to contact me/us for more information.

Btw, I added the 'arbitrary value' example to answer your last question: No, HbO is not always smaller than HbT, because you can also subtract an arbitrary large offset from HbR, making the 'relative' change in HbR still valid, but leading to HbT being smaller than HbO. Most devices and/or toolboxes actually do baseline with respect to the first recorded sample.

False sentence is totally wrong i.e. He was arrested for giving false information on his application for a passport. Incongruous mean unsuitable, inappropriate, inconsistent, inharmonious, incompatible, conflicting, discordant. For example: It seems incongruous to have an out-of-shape and overweight editor of a fitness magazine.

There are quite new methods from our lab (Machine Learning Department, TU Berlin) that approach the extraction of Information on an early source Separation Level:

For more info see Dähne et al 2015 (Multivariate Machine Learning Methods for Fusing Multimodal  Functional Neuroimaging Data)

and Dähne et al 2013 (Integration of Multivariate Data Streams With Bandpower Signals)

Best

Alex

Dear Laipeng, 

Check this paper out if it gives you some ideas:

Yes, this is 4000mV at detector (due to light intensity) that goes beyond the Analog to Digital (A/D) converter limit and will results in saturation for higher values. You need to adjust LED current and brightness to lower the light intensity at the detector.

To check whether it is broken or it works, the best form is to used an optical phantom for which you know the properties in advance. They are "easy" to build. Also, some signal output are characteristic of malfunction; e.g. saturation of one wavelength leads to signal mirroring, saturation of 2 wavelengths leads to a flat response, etc. To give you a better answer, you might need to be more precise about what you are observing which is leading you to think that the device is not working.

Regarding parameters of applicability, I know it is a bit of self publicity which is perhaps inappropriate -hope you don't mind-, but you may be interested in our;

Orihuela-Espina F, Leff DR, James DR, Darzi AW, Yang GZ. "Quality control and assurance in functional near infrared spectroscopy (fNIRS) experimentation." Phys Med Biol. 2010 Jul 7;55(13):3701-24. doi: 10.1088/0031-9155/55/13/009.

 

Yes, I would recommend 5 for LED current and 1 for gain as a start. Also, looking at light intensity data is important, COBI Studio saves raw light intensity data in nir file and raw hemoglobin data in oxy file. I would be happy to discuss this offline and through telecon for faster progress.

I think the answer is dependent also on the task given to the child. There are some useful articles. 

Haaland K.Y., Prestopnik J.L., Knight R.T., Lee R.R. Hemispheric asymmetries for kinematic and positional aspects of reaching. Brain. 2004. 127(5): 1145-1158.

Harrington D.L., Haaland K.Y. Hemispheric specialization for motor sequencing: abnormalities in levels of programming. Neuropsychology. 1991. 29(2): 147-163.

 I think Mr.Zarei wants to be familiar with a  standard method to calibrate fNIRS, what do you think, is there any method for this calibration? for example, a homogeneous phantom can be used for this calibration?

thanks Mr. Horschig,

It depends on the frequency of the signal output, which hasn't been mentioned. The centre frequency of the bandpass filter should be the same as the frequency of the signal. The sampling rate should be at least double the frequency of the signal. However, as I have not seen your experiment, your mileage may vary.

The link will explain sampling rate and why it needs to be 2.7 times greater than the frequncy of the signal. https://en.m.wikipedia.org/wiki/Sampling_(signal_processing)

The answer really depends on the time horizon that you want to consider.  I would say that within the next 5-10 years, minimally, surface brain recordings will continue to play pretty limited roles.  Among the issues that continue to plague these technologies are both spatial and temporal sampling/resolution issues - especially given the problems of the skull and brain as, essentially, large biological volume conductors. The amount of computational effort required to clean up and de-mix source signals means that real-time applications will continue to be very limited. Mostly, one can expect only very general and non-specific metrics to be most useful for near-term real-time systems (basic spectral power and simple event-related measures are most likely to continue to hold their place as mainstays).  Over longer time horizons, of course, computational methods will become more efficient both in formulation as well as in researchers and developers learning what information is NOT needed to be processed.  Improvements in hardware/firmware/etc also will continue to provide better opportunities for real-time applications.  However, the final hurdle of understanding exactly what information is contained within these types of neural measures is going to unfold over  a much, much longer time course. I personally think that the highest payoff areas will continue to be the monitoring of much lower-dimensional signals of peripheral psychophysiology for some time to come . . . and, let's face it, there is no Tony Stark and Iron Man is not coming in any near future that I can see . . . 

Yes, using the modified Beer-Lambert law. You can see the math and its easy explanation in the following papers (as I can not type math in the comment). Let me know in case you require further help.

Andreas that sounds like a very good idea! We all spend most of our time in applying for expensive resources which are often used only for one or two protocols. Exchange of these resources would be a tremendous step forward. In fact we do even have the duty to share our resources which are all provided by the tax payers. We are doing exactly this with our fMRI equipment: colleagues come to our lab to work on their questions and we provide support for that. However, these colleagues are always people we have previously worked with in other Universities. It should also be possible without the personal link. It would be extremely fruitful to share TMS-labs or EEG-resources. It would be great to build up a network of resources and researchers who are happy to exchange. Mathias, thanks for the great link to "EuroBioImaging". That seems to be a perfect realization of the idea Andreas posted!

Matthew's suggestion is good. You can also look at the examples of how Homer2 does it at homer-fnirs.org . And also this paper

Huppert, Theodore J, Solomon G Diamond, Maria A Franceschini, and David A Boas. 2009. “HomER: a Review of Time-Series Analysis Methods for Near-Infrared Spectroscopy of the Brain..” Applied Optics 48 (10): D280–98.

Hi,

There is an error in data filtering section. It is not possible to synthesize a Butterworth band-pass filter of such a low order. You can use an optimal elliptic filter, for example:

Wn = [0.01 0.1];

Fs = 1.53; Fn = Fs/2;

N = 5; Rp = .5; Rs = 50;

[b,a] = ellip(N,Rp,Rs,Wn/Fn);

You can see the magnitude and phase response of the filter using Matlab's fvtool:

fvtool(b,a);

Best regards,

Rosen

Epskamp, S., Cramer, A.O.J., Waldorp, L.J., Schmittmann, V.D. and Borsboom, D. (2012) qgraph: Network Visualizations of Relationshipsin Psychometric Data. Journal of Statistical Software, 48(4), 1-18.

My pleasure Kyungu. You may also contact Oldemiro Rego (orego@uac.pt) who works straight with Alfredo Borba in your field.

All the best,

JP