Wednesday, August 15, 2018

ECoG Post-doc opportunities with Nitin Tandon, Stan Dehaene, Nathan Crone, Xaq Pitkow & me


Postdoctoral research positions are available in the lab of Nitin Tandon at Houston. This position is funded by a BRAIN Initiative U01 grant funded project that uses electro-corticographic (ECoG) recordings and fMRI on a large cohort (n=80) to evaluate psycholinguistic models of reading and speech production to create network level representation of language. Collaborators on the project include Greg Hickok, Stanislas Dehaene, Nathan Crone and Xaq Pitkow; the post-doc will benefit from a close interaction with these experts in the fields of reading, semantics, speech production and computational neuroscience.
The selected individuals are expected to be a highly motivated, team players with a passion to study cognitive processes using imaging, direct recordings and closed-loop cortical stimulation in humans. They will be responsible for
1)    data collection in the epilepsy monitoring unit and in the MRI scanner
2)    ECoG data analysis using an analysis pipelines existent in the lab and via the development of innovative analytic strategies including machine learning and AI approaches
3)    data presentation at conferences and manuscript preparation.
The selected individual must have a Ph.D. in one or more of the following: neuroscience, psychology, cognitive science, mathematics, electrical engineering or computer science. Previous experience in neural time series data analysis, functional imaging studies of language, or studies of speech production are desirable – but not crucial. They must possess the ability to independently code in any or all of the following: MATLAB, R or python.

Given the multiple unpredictable variables and privacy issues around data collection in human patients, the individual must possess high ethical and professionalism standards, be able to adapt to a changing environment, reorganize schedules dynamically, and work with short deadlines. A strong publication record and excellent academic credentials are highly desirable.


Wednesday, August 8, 2018

The neural basis of word deafness: a frequency analysis

There are two competing theories regarding the neural basis of word deafness.  Empirically we know that it can occur following unilateral left or bilateral lesions in and around auditory cortex (STG).  

The unilateral theory says that the relevant speech perception network is left dominant.  Unilateral lesions cause word deafness by simultaneously disconnecting acoustic input from ascending auditory pathway AND by disconnecting callosal fibers from the intact right hemisphere.  Bilateral lesions do the same but clip the right hemisphere inputs at their source. 

The bilateral theory says that the relevant speech perception network is bilateral.  Bilateral lesions disrupt the speech perception network, therefore causing word deafness.  Unilateral cases are anomalies, perhaps reflecting atypically strong left dominant language organization.  

Here's an argument in favor of the bilateral account based on probability distributions.  Note that there are several simplifying assumptions here, but I believe these bias things in favor of the unilateral theory.

According to the CDC (as of September 6, 2017), 795,000 people in the United States have a stroke every year, 610,000 of which are first time strokes.  From these numbers, and assuming a 50/50 chance of having a left or right hemisphere stroke, we can estimate the following in a given year:

·      Number of people with first-time left hemisphere strokes: 610,000 * .5 = 305,000
·      Number of people who have a second (or 3rd or 4th…) stroke: 795,000 – 610,000 = 185,000
·      Number of people whose second stroke is in the opposite hemisphere from their first stroke: 185,000 * .5 = 92,500
·      Number of people with either a left hemisphere stroke or a bilateral stroke: 305,000 + 92,500 = 397,500

Now, with the three bolded numbers, we can calculate expected proportion of a single left hemisphere stroke compared with a bilateral stroke in this sample of 397,500 with one or the other:

·      Proportion of single left hemisphere strokes: 305,000/397,500 = .767
·      Proportion of bilateral strokes: 92,500/397,500 = .233

With these proportions we can estimate the expected frequency of word deaf cases with unilateral versus bilateral strokes, assuming an equal frequency of occurrence for either etiology for a given sample size.  Here we use a sample of 59 cases with confirmed lesions, as discussed by Poeppel (2001).  The expected frequencies are 45.3 (59 * .767) cases with unilateral left hemisphere lesions and 13.73 (59 * .233) cases with bilateral lesions. The actual frequencies are 17 cases with left hemisphere lesions and 42 cases with bilateral lesions. We can then plug these numbers into a binomial probability calculation--assuming a biased coin flip with the bias equal to .767 in favor of unilateral--to determine the probability of getting such an actual outcome. When we do this, we find that the observed outcome is extremely unlikely and would be expected to occur by chance only 1 time in 1 million samples (p = 0.000001). 

This seems like pretty strong evidence in favor of the bilateral theory.