Reader Comments
Post a new comment on this article
Post Your Discussion Comment
Please follow our guidelines for comments and review our competing interests policy. Comments that do not conform to our guidelines will be promptly removed and the user account disabled. The following must be avoided:
- Remarks that could be interpreted as allegations of misconduct
- Unsupported assertions or statements
- Inflammatory or insulting language
Thank You!
Thank you for taking the time to flag this posting; we review flagged postings on a regular basis.
closeVery interesting work!
Posted by AnimeshRay on 23 Oct 2014 at 02:29 GMT
This is a beautiful work. I am not a neuroscientist, so I do not obviously understand many of the points, but upon a quick read I have a few questions.
First, I see that you took pains to ensure that both groups of subjects were "wakeful" using I suppose criteria that are considered valid. However, what is the evidence--how confident are we--that the criteria of the 'signals of wakefulness' by which one can characterize that state do indeed hold for the brains of DoC patients? Suppose one were to conduct the same analysis on two groups of healthy volunteers, one 'awake' and another 'asleep'. Might one find similar differences in the connectivity patterns? Would the difference between 'awake' and 'sleeping' brains be as striking? Could there be features of 'sleep' embedded within the DoC brain signals, and perhaps sleep/awake data could potentially disentangle that component?
Second, since most (all) definitions of modularity are arbitrary. One generally uses a particular modular decomposition and tries to figure whether that particular decomposition has some functional significance, by looking for statistical enrichment of some attributes (in this case, for e.g., it could have been functional groups of cortical neurons). I don't see a discussion of this issue in the paper. Do the Louvain modules represent functionally significant modules? What if you used a different modular decomposition, such as Markovian clusters or Newman's Q? Do you see significant differences in the 'modular span' with a different modular decomposition compared to Louvain module?
Nonetheless, this is a wonderful paper, and my Facebook friends are all 'liking' it! :-)
RE: Very interesting work!
srivaschennu replied to AnimeshRay on 23 Oct 2014 at 21:46 GMT
Dear Animesh,
Thank you for your interest, in addition to the detailed and insightful comments. In response to the key points you raised:
1. Arousal (or vigilance) varies in complex ways in vegetative patients (Landsness et al., 2011; Brain), and indeed, it is often non-trivial to accurately assess true levels of arousal using behaviour. However, as we show using eye-movement activity, there was no evidence that our patients were systematically falling asleep. Hence, though they were unlikely to be asleep throughout, ongoing and rapid fluctuations of arousal in patients during the recording warrants careful interpretation of the limitations and generalisability of these measures.
2. One would certainly expect changes in connectivity during sleep onset and during slow-wave sleep, and there are several studies that have investigated this (Langheim et al., 2011; Tagliazucchi et al., 2013; Ferri et al., 2007, 2008). These lead us to think that the changes in connectivity that distinguish sleep stages from normal wakefulness would be distinct from the changes in connectivity described here between normal wakefulness and disorders of consciousness. Also, the same distinction applies to the comparison of normal wakefulness to propofol sedation, for example.
3. Your suggestion of exploring 'latent' sleep-related connectivity markers in disorders of consciousness patients is certainly very valuable, and something to follow up in the future!
4. The modular decomposition applied here was by definition oblivious to the source of the connectivity data, and hence just relied on strength of connectivity to heuristically identify what it thought to be modules of strongly interconnected nodes. When these modules were overlaid on the head, it ended up highlighting frontoparietal linkages in connectivity within the alpha band in healthy adults (and some patients), and that was certainly interesting. In this sense, not biasing the modularity algorithms by necessitating a brain-inspired functional significance allowed us to use graph theory to make a pure 'topological inference' and later ask if that topological inference actually corresponded to something topographically sensible. Hope that makes sense. We do make this point in the paper, in linking topology to topography via modular span. But your point about identifying modular decompositions that try to maximise the 'statistical enrichment' of some brain-derived attributes is novel, and I think this would yield more illuminating outcomes with fMRI data, where there is spatial detail available to mine.
5. Finally, no we haven't yet tried to use other modular decomposition algorithms; again another methodologically important avenue to explore.
Many thanks again for your comments and suggestions. Glad you liked the paper!