A new era of brain cartography, powered by neuroimaging and machine learning

– M GLASSER, D VAN ESSEN/WASHINGTON UNIVERSITY

Human brain mapped in unprece­dent­ed detail (Nature):

“Think of a spin­ning globe and the patch­work of coun­tries it depicts: such maps help us to under­stand where we are, and that nations dif­fer from one anoth­er. Now, neu­ro­sci­en­tists have chart­ed an equiv­a­lent map of the brain’s out­er­most lay­er — the cere­bral cor­tex — sub­di­vid­ing each hemi­sphere’s moun­tain- and val­ley-like folds into 180 sep­a­rate parcels.

Nine­ty-sev­en of these areas have nev­er pre­vi­ous­ly been described, despite show­ing clear dif­fer­ences in struc­ture, func­tion and con­nec­tiv­i­ty from their neigh­bours. The new brain map is pub­lished today in Nature…

“While the focus of this work was on cre­at­ing a beau­ti­ful, reli­able, aver­age brain tem­plate, it real­ly opens up the pos­si­bil­i­ty to fur­ther explore the unique inter­sec­tion of indi­vid­ual tal­ents with intel­lec­tu­al and cre­ative abil­i­ties — the things that make us unique­ly human,” says Rex Jung, a neu­ropsy­chol­o­gist at the Uni­ver­si­ty of New Mex­i­co in Albuquerque.

Study: A mul­ti-modal par­cel­la­tion of human cere­bral cor­tex (Nature; opens PDF)

• Abstract: Under­stand­ing the amaz­ing­ly com­plex human cere­bral cor­tex requires a map (or par­cel­la­tion) of its major sub­di­vi­sions, known as cor­ti­cal areas. Mak­ing an accu­rate are­al map has been a cen­tu­ry-old objec­tive in neu­ro­science. Using mul­ti-modal mag­net­ic res­o­nance images from the Human Con­nec­tome Project (HCP) and an objec­tive semi-auto­mat­ed neu­roanatom­i­cal approach, we delin­eat­ed 180 areas per hemi­sphere bound­ed by sharp changes in cor­ti­cal archi­tec­ture, func­tion, con­nec­tiv­i­ty, and/or topog­ra­phy in a pre­cise­ly aligned group aver­age of 210 healthy young adults. We char­ac­ter­ized 97 new areas and 83 areas pre­vi­ous­ly report­ed using post-mortem microscopy or oth­er spe­cial­ized study-spe­cif­ic approach­es. To enable auto­mat­ed delin­eation and iden­ti­fi­ca­tion of these areas in new HCP sub­jects and in future stud­ies, we trained a machine-learn­ing clas­si­fi­er to rec­og­nize the mul­ti-modal ‘fin­ger­print’ of each cor­ti­cal area. This clas­si­fi­er detect­ed the pres­ence of 96.6% of the cor­ti­cal areas in new sub­jects, repli­cat­ed the group par­cel­la­tion, and could cor­rect­ly locate areas in indi­vid­u­als with atyp­i­cal par­cel­la­tions. The freely avail­able par­cel­la­tion and clas­si­fi­er will enable sub­stan­tial­ly improved neu­roanatom­i­cal pre­ci­sion for stud­ies of the struc­tur­al and func­tion­al orga­ni­za­tion of human cere­bral cor­tex and its vari­a­tion across indi­vid­u­als and in devel­op­ment, aging, and disease.

To learn more:

• To Be (Your Con­nec­tome), or Not to Be (Your Genome)
• Human Con­nec­tome Project