Deep learning model built on neuroimaging data identifies “Brain Age Gaps” as markers of Alzheimer’s disease (AD)

Brain Age Gap is a Com­pos­ite Bio­mark­er for Demen­tia Pathol­o­gy or Sever­i­ty (GEN):

Mayo Clin­ic sci­en­tists have devel­oped a com­pu­ta­tion­al mod­el that pre­dicts brain age using a large col­lec­tion of neu­roimag­ing data obtained using FDG-PET (flu­o­rodeoxyglu­cose positron emis­sion tomog­ra­phy) and struc­tur­al MRI (mag­net­ic res­o­nance imag­ing). The deep learn­ing-based mod­el tests the rela­tion­ship between brain age gaps in var­i­ous forms of demen­tia, includ­ing mild cog­ni­tive impair­ment (MCI), Alzheimer’s dis­ease (AD), fron­totem­po­ral demen­tia (FTD), and Lewy body demen­tia (LBD), as well as in nor­mal brains.

… “The abil­i­ty for deep learn­ing to accu­rate­ly pre­dict age based on brain imag­ing data has been known for some time. How­ev­er, look­ing at brain age gap or the dif­fer­ence between pre­dict­ed and actu­al age, has been thought to have the poten­tial to be uti­lized as a bio­mark­er. Oth­ers have argued that such a brain age gap is only able to mark treat­ment-lev­el bio­log­i­cal dif­fer­ences and is unable to track changes in state and there­fore should not be inter­pret­ed as accel­er­at­ed brain aging,” (Senior author of the study, Dr. David) Jones said. “The main find­ing of our study is that we could indeed find evi­dence that high brain age gap is behav­ing as an accel­er­at­ed brain aging biomarker.”

The Study:

Deep learn­ing-based brain age pre­dic­tion in nor­mal aging and demen­tia (Nature Aging).

Abstract: Brain aging is accom­pa­nied by pat­terns of func­tion­al and struc­tur­al change. Alzheimer’s dis­ease (AD), a rep­re­sen­ta­tive neu­rode­gen­er­a­tive dis­ease, has been linked to accel­er­at­ed brain aging. Here, we devel­oped a deep learn­ing-based brain age pre­dic­tion mod­el using a large col­lec­tion of flu­o­rodeoxyglu­cose positron emis­sion tomog­ra­phy and struc­tur­al mag­net­ic res­o­nance imag­ing and test­ed how the brain age gap relates to degen­er­a­tive syn­dromes includ­ing mild cog­ni­tive impair­ment, AD, fron­totem­po­ral demen­tia and Lewy body demen­tia. Occlu­sion analy­sis, per­formed to facil­i­tate the inter­pre­ta­tion of the mod­el, revealed that the mod­el learns an age- and modal­i­ty-spe­cif­ic pat­tern of brain aging. The ele­vat­ed brain age gap was high­ly cor­re­lat­ed with cog­ni­tive impair­ment and the AD bio­mark­er. The high­er gap also showed a lon­gi­tu­di­nal pre­dic­tive nature across clin­i­cal cat­e­gories, includ­ing cog­ni­tive­ly unim­paired indi­vid­u­als who con­vert­ed to a clin­i­cal stage. How­ev­er, regions gen­er­at­ing brain age gaps were dif­fer­ent for each diag­nos­tic group of which the AD con­tin­u­um showed sim­i­lar pat­terns to nor­mal aging.

News in Context:

• Geisinger and Eisai to test real-world valid­i­ty of AI-pow­ered Pas­sive Dig­i­tal Mark­er (PDM) in detect­ing ear­ly cog­ni­tive impair­ment and dementia
• Bea­con Biosig­nals rais­es $27M to scale EEG, AI-based neu­ro­bio­mark­er dis­cov­ery platform