U.S. Army develops novel way to analyze brain imaging data and shape emerging non-invasive neurotechnology

by

Aggre­gate dis­tri­b­u­tion of cor­ti­cal brain-wave activ­i­ty, orga­nized by stan­dard fre­quen­cy bands, across a range of depths. Cred­it: U.S. Army

Army devel­ops big data approach to neu­ro­science (press release):

“A big data approach to neu­ro­science promis­es to sig­nif­i­cant­ly improve our under­stand­ing of the rela­tion­ship between brain activ­i­ty and performance.

To date, there have been rel­a­tive­ly few attempts to use a big-data approach with­in the emerg­ing field of neu­rotech­nol­o­gy. In this field, the few attempts at meta-analy­sis (analy­sis across mul­ti­ple stud­ies) com­bine only the results from indi­vid­ual stud­ies rather than the raw data. A new study is one of the first to com­bine data across a diverse set of exper­i­ments to iden­ti­fy pat­terns of brain activ­i­ty that are com­mon across tasks and people.

The Army in par­tic­u­lar is inter­est­ed in how the cog­ni­tive state of Sol­diers can affect their per­for­mance dur­ing a mis­sion. If you can under­stand the brain, you can pre­dict and even enhance cog­ni­tive performance.

Researchers from the U.S. Army Com­bat Capa­bil­i­ties Devel­op­ment Com­mand’s Army Research Lab­o­ra­to­ry teamed with the Uni­ver­si­ty of Texas at San Anto­nio and Intheon Labs to devel­op a first-of-its-kind mega-analy­sis of brain imag­ing data–in this case elec­troen­cephalog­ra­phy, or EEG.

In the two-part paper, they aggre­gate the raw data from 17 indi­vid­ual stud­ies, col­lect­ed at six dif­fer­ent loca­tions, into a sin­gle ana­lyt­i­cal frame­work, with their find­ings pub­lished in a series of two papers in the jour­nal Neu­roIm­age. The indi­vid­ual stud­ies includ­ed in this analy­sis encom­pass a diverse set of tasks such sim­u­lat­ed dri­ving and visu­al search.”

The Study:

Auto­mat­ed EEG mega-analy­sis I: Spec­tral and ampli­tude char­ac­ter­is­tics across stud­ies (Neu­roIm­age)

  • From the abstract: Sig­nif­i­cant achieve­ments have been made in the fMRI field by pool­ing sta­tis­ti­cal results from mul­ti­ple stud­ies (meta-analy­sis). More recent­ly, fMRI stan­dard­iza­tion efforts have focused on enabling the joint analy­sis of raw fMRI data across stud­ies (mega-analy­sis), with the hope of achiev­ing more detailed insights. How­ev­er, it has not been clear if such analy­ses in the EEG field are pos­si­ble or equal­ly fruit­ful. Here we present the results of a large-scale EEG mega-analy­sis using 18 stud­ies from six sites rep­re­sent­ing sev­er­al dif­fer­ent exper­i­men­tal par­a­digms. We demon­strate that when meta-data are con­sis­tent across stud­ies, both chan­nel-lev­el and source-lev­el EEG mega-analy­sis are pos­si­ble and can pro­vide insights unavail­able in sin­gle studies…In a com­pan­ion paper, we apply mega-analy­sis to assess com­mon­al­i­ties in event-relat­ed EEG fea­tures across stud­ies. The con­tin­u­ous raw and pre­processed data used in this analy­sis are avail­able through the Dat­a­Cat­a­log at https://cancta.net.

Auto­mat­ed EEG mega-analy­sis II: Cog­ni­tive aspects of event relat­ed fea­tures (Neu­roIm­age)

  • From the abstract: We present the results of a large-scale analy­sis of event-relat­ed respons­es based on raw EEG data from 17 stud­ies per­formed at six exper­i­men­tal sites asso­ci­at­ed with four dif­fer­ent insti­tu­tions. The analy­sis cor­pus rep­re­sents 1,155 record­ings con­tain­ing approx­i­mate­ly 7.8 mil­lion event instances acquired under sev­er­al dif­fer­ent exper­i­men­tal par­a­digms. Such large-scale analy­sis is pred­i­cat­ed on con­sis­tent data orga­ni­za­tion and event anno­ta­tion as well as an effec­tive auto­mat­ed pre­pro­cess­ing pipeline to trans­form raw EEG into a form suit­able for com­par­a­tive analysis…This work demon­strates that EEG mega-analy­sis (pool­ing of raw data across stud­ies) can enable inves­ti­ga­tions of brain dynam­ics in a more gen­er­al­ized fash­ion than sin­gle stud­ies afford. A com­pan­ion paper com­ple­ments this event-based analy­sis by address­ing com­mon­al­i­ty of the time and fre­quen­cy sta­tis­ti­cal prop­er­ties of EEG across stud­ies at the chan­nel and dipole level.

News in Context: