[Homer-users] Odd HRF shape: possible explanations?

[Luca Filippin]

Thanks Felix for your comments and for the bibliography.

We carefully selected the best babies so to reduce the impact of the
movement artifacts.
We applies a standard lowpass filtering (cut frequency at 1Hz) & signal
correction. Homer removes from the analysis the stimuli around which the
signals are too compromised.  Unfortunately we didn't measure any
physiological confounds: heart beat, respiration etc.

Best,
L.


On Sun, Mar 12, 2017 at 10:26 PM, Scholkmann Felix <Felix.Scholkmann at usz.ch>
wrote:

> Dear Luca,
>
>
> From my own fNIRS studies done with neonates and from published papers I
> know that there are several factors that are causing the HRF of neonates
> being different compared to the HRF of adults:
>
>
> (1) Normaly, the amount of movement artifacts in fNIRS signals
> from neonates is larger compared to recordings from adults. This, in
> combination with the filtering, can cause distortions of the shape of
> the HRF. In oder to prevent this, all trials that are heavily contaminated
> with artifacts should be rejected. If the artifacts are not that severe the
> signal can be reconstructed. There are several methods available. For
> example, the algorithm published in 2010 [1] (I can send you the Matlab
> code if you need it), or other methods as implemented in NIRS-SPM and Homer.
>
>
> (2) According to findings from Chen et al. [2] the neuro-vascular
> coupling comprises two components: a fast and a slow mechanical change in
> the endothelium. The fast component (due to endothelial hyperpolarization)
> has a high propagation speed (about 2 mm/s), the slow one (due to calcium
> wave-associated dilation) propagated much slower. These two compoents form
> the shape of the HRF. In adults, these two mechanism work quite well
> - which is not the case in neonates. In neonates, one mechanis is well
> developed, the other not - as shown by Kozberg et al. [3]:
>
> *"Our data suggest that there are at least two vascular mechanisms at work
> in normal neurovascular coupling: (**i**) a global, delayed constriction
> process, present from early development and (**ii**) a localized initial
> hyperemia that develops gradually and is not in place at birth."*
>
> So, in neonates (depending on the age), the neuro-vascular coupling is
> different and will produce different HRFs than in adults.
>
>
> (3) fNIRS signals are contaminated with physiological signals coming from
> the extracerebral compartment (skalp blood flow) and
> systemic physiology affecting the global cerebral blood flow. See the
> review of 2016: [4]. CW-NIRS systems with only one source-detector system
> are susceptible to these artifacts.
>
> If there is a global changes in blood pressure, the fNIRS signals will be
> affeced. This is especially true for neonates - as nicely shown by Kozberg
> et al. [3]:
>
> *"We also observed that stimulus-evoked systemic blood pressure increases
> can cause apparent adult-like positive BOLD cortical responses in neonates
> consistent with immature autoregulation. We conclude that the parallel
> development and complex interplay of autoregulation, neurovascular
> coupling, and neuronal function in the developing brain should be carefully
> considered when interpreting hemodynamic measures such as fMRI BOLD in
> infants and children."*
>
> See also the figures in this paper that visualize the problem.
>
>
> (4) And finally, the shape of the HRF depends on the type of filtering
> applied. The main parameters are the lower and upper cutoff-frequencies of
> the bandpass-filter and the filter type (as well as the order of the
> filter). Changing these parameters will change the shapes. In addition, the
> shapes depend also on the choise of either using the mean or the median
> for calculating the block-averages. I recommend using the median.
>
>
> In summary, there are many factors involved causing "abnormal" HRF shapes
> in fNIRS recordings of neonates. It's a mixture of physiology and
> technology that determines how the shapes will look. The best was is to
> report precisely all the signal processing steps performed and to conduct
> additional physiological measurements (e.g. hear rate, SaO2, respiration
> rate, etc.) during the fNIRS experiments in order to allow a correct
> physiological interpretation of the fNIRS signals.
>
>
> Kind regards,
>
> Felix
>
>
> [1] http://iopscience.iop.org/article/10.1088/0967-3334/31/5/004/meta
>
> [2] http://jaha.ahajournals.org/content/3/3/e000787
>
> [3] https://www.ncbi.nlm.nih.gov/pubmed/23426630
>
> [4] http://neurophotonics.spiedigitallibrary.org/
> article.aspx?articleid=2502725
>
>
>
>
>
>
> ------------------------------
> *Von:* homer-users-bounces at nmr.mgh.harvard.edu <
> homer-users-bounces at nmr.mgh.harvard.edu> im Auftrag von Cecile Issard <
> cecile.issard at etu.parisdescartes.fr>
> *Gesendet:* Sonntag, 12. März 2017 17:10
> *An:* homer-users at nmr.mgh.harvard.edu
> *Betreff:* Re: [Homer-users] Odd HRF shape: possible explanations?
>
>
> Hello Luca,
>
> What are the 40 s you average? Only one block? Or a block + silence? It
> seems to me that 50 to 80 ms is very short in between blocks.
>
> Best,
>
> Cécile Issard
> PhD student
> Laboratoire Psychologie de la Perception
> Université Paris Descartes- CNRS
> Paris, France
> Le 11 mars 2017 5:06 PM, Luca Filippin <luca.filippin at gmail.com> a écrit :
>
> Hi,
>    we've been using Homer2 to analyze some data collected on newborns and
> we get some puzzling HRF shapes as result, averaged over a 40 secs temporal
> window The function has in some cases multiple humps: it rises then
> descends a bit and then rises again, descends again and so on, 2 or more
> times. In some other cases it simply rises and doesn't fall at all.
>
> Would you have any explanations for that?
>
> We believe that it has something to do with the stimuli design.
>
> Our stimuli are sounds of variable length between 18 and 22 secs, roughly.
> Each sound is made of block of N repeated short sound segments (for
> example: BI), interleaved with other blocks of M sound segments of the same
> type BI, but shorter duration:
>
> N x BI (pause) M x shorter BI  (variable pause) N x BI (variable pause) M
> x shorter BI  (variable pause) ...
>
> where "variable pause" is a silent segment of duration between 50 and 80
> ms.
>
> We wonder if  there's a way to do the analysis so to disentagle the
> contribution of each block of sound segment within a sound stimulus. We
> play different sounds during the experiments: each sounds onset is
> separated by roughly 50 seconds. The problem would be perhaps, the
> estimation of the baseline.
>
> Many thanks for your help,
> Luca
>
>
>
>
>
> _______________________________________________
> Homer-users mailing list
> Homer-users at nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/homer-users
>
>
> The information in this e-mail is intended only for the person to whom it
> is
> addressed. If you believe this e-mail was sent to you in error and the
> e-mail
> contains patient information, please contact the Partners Compliance
> HelpLine at
> http://www.partners.org/complianceline . If the e-mail was sent to you in
> error
> but does not contain patient information, please contact the sender and
> properly
> dispose of the e-mail.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/homer-users/attachments/20170313/a44f27dd/attachment.html