A1-PaulIvanov

From CS 294-10 Visualization Sp10

Jump to: navigation, search

Good

Yuval-Greenberg S., Tomer O., Keren A., Nelken I., Deouell L. Transient Induced Gamma-Band Response in EEG as a Manifestation of Miniature Saccades. Neuron, Volume 58, Issue 3, Pages 429-441

Image:Yuval-Greenberg_Neuron_2008.png

The roles of brain activity at different temporal frequencies is an active research area in neuroscience. Occipital gamma-band activity, roughly 30-100Hz, is thought to be related to the object recognition and visual attention. This paper illustrates that, at least in the case of electroencephalography (EEG), what was previously thought to be gamma band activity is an artifact caused by fixational eye-movements (microsaccades).

Subplots A and B show the saccades and gamma-band EEG peaks, respectively, as both trial raster plots, and rate histograms collapsed across trials. In both subplots, the red line represents saccade rate, which is a nice thread to tie these two subplots together. There's a compelling suppression of both saccade and gamma-band activity 100-200ms, followed by their concurrent increase 200-300ms.

Further insight comes from reordering the trials based on how far into the trial the saccade occurred, if one occurred at all. Trials without microsaccades (C:top, above white line) had no gamma activity (C:bottom, above white line). In the trials which had them, the onset of microsaccades were immediately followed by a gamma band response (D, ordered by saccade onsets, saccade onsets in black). In this case, it turns out that the response was actually *very* broadband. The reason it appeared to be in the gamma band was because it was filtered to just the gamma band.


Bad

Felleman, D.J. and Van Essen, D.C. (1991) Distributed hierarchical processing in primate visual cortex. Cerebral Cortex, 1: 1-47

Image:Felleman_CC_1991.png

This is a very frequently shown visualization of the organization of the visual system. Part of the problem is that there are too many links between every visual area to make much sense out of it. It's often hard to follow connections because there are so many ambiguous intersections. Additional confusion comes from the fact that some of the connections are red instead of black.

Additionally, a crucial piece of information absent from this visualization is information about the ABSENCE of links between different areas. Just because there isn't a link between two areas, does not necessarily mean that such a link does not exist. Only a portion of pathways has been explicitly tested and found to be absent. There's an ambiguity: an absent link can either mean that there *is* experimental evidence supporting the absence of a connection between areas, but it can also mean that there simply been *no* experiments performed to check the existence of a connection between two areas.

Table 3 in the same paper is a connectivity matrix which serves as a much more interpretable representation of the data. Not only does it differentiate between pathways not carefully tested (as blanks) and pathways tested and found to be absent (as dots '.'), it further differentiates between present connections identified in full length manuscripts (big '+') and only in abstracts and unpublished data (little '+'), and identifies nonreciprocal or one-way pathways (as 'NR'), as well as pathways whose existence is uncertain due to conflicting reports in literature ('?')



[add comment]
Personal tools