AI doctor

Researchers use artificial intelligence to uncover the cellular origins of Alzheimer’s disease and other cognitive disorders

Researchers have used arti­fi­cial intel­li­gence tech­niques to exam­ine struc­tur­al and cel­lu­lar fea­tures of human brain tis­sues to help deter­mine the caus­es of Alzheimer’s dis­ease and oth­er relat­ed disorders.

Instead of using tra­di­tion­al mark­ers like amy­loid plaques, the Mount Sinai research team found that study­ing the caus­es of cog­ni­tive impair­ment using an unbi­ased AI-based tech­nique revealed unex­pect­ed micro­scop­ic abnor­mal­i­ties that might pre­dict the pres­ence of cog­ni­tive impair­ment. On Sep­tem­ber 20, these find­ings were pub­lished in the jour­nal Acta Neu­ropatho­log­i­ca Communications.

Stat­ing that AI rep­re­sents an entire­ly new par­a­digm for study­ing demen­tia and will have a trans­for­ma­tive effect on research into com­plex brain dis­eases, espe­cial­ly Alzheimer’s dis­ease, co-cor­re­spond­ing author John Crary, MD, PhD, Pro­fes­sor of Pathol­o­gy, Mol­e­c­u­lar and Cell-Based Med­i­cine, Neu­ro­science, and Arti­fi­cial Intel­li­gence and Human Health, at the Icahn School of Med­i­cine at Mount Sinai, said, “The deep learn­ing approach was applied to the pre­dic­tion of cog­ni­tive impair­ment, a chal­leng­ing prob­lem for which no cur­rent human-per­formed histopatho­log­ic diag­nos­tic tool exists.”

The medi­al tem­po­ral lobe and frontal cor­tex were the two brain regions whose under­ly­ing archi­tec­ture and cel­lu­lar fea­tures the research team iden­ti­fied and ana­lyzed. The researchers used a weak­ly super­vised deep learn­ing algo­rithm to assess slide images of human brain autop­sy tis­sues from a group of more than 700 elder­ly donors in order to pre­dict the pres­ence or absence of cog­ni­tive impair­ment in an attempt to improve the stan­dard for post­mortem brain assess­ment in order to iden­ti­fy signs of diseases.

Related Readings:

In a super­vised learn­ing envi­ron­ment, the weak­ly super­vised deep learn­ing approach can han­dle noisy, lim­it­ed, or impre­cise sources to pro­vide sig­nals for label­ing sub­stan­tial amounts of train­ing data. The amount of myelin, the pro­tec­tive lay­er around brain nerves, is mea­sured using the Lux­ol rapid blue stain­ing, and this deep learn­ing mod­el was used to iden­ti­fy a decrease in the quan­ti­ty of myelin.

The white mat­ter, which is involved in learn­ing and brain func­tions, showed a con­cen­tra­tion of myelin stain­ing that was decreas­ing in amount, dis­persed in a non-uni­form man­ner across the tis­sue, and asso­ci­at­ed with cog­ni­tive impair­ment. The accu­ra­cy of the two sets of mod­els that the researchers devel­oped and used to pre­dict the pres­ence of cog­ni­tive impair­ment was bet­ter than that of guessing.

Accord­ing to their find­ings, the dimin­ished stain­ing inten­si­ty in cer­tain brain regions iden­ti­fied by AI may pro­vide a scaleable plat­form to assess the pres­ence of brain impair­ment in oth­er relat­ed disorders.

The method­ol­o­gy estab­lish­es the frame­work for fur­ther research, which may involve using arti­fi­cial intel­li­gence mod­els on a big­ger scale and fur­ther dis­sect­ing the algo­rithms to improve their reli­a­bil­i­ty and accu­ra­cy. The group said that the ulti­mate aim of this neu­ropatho­log­ic research pro­gram is to cre­ate bet­ter ther­a­peu­tic and diag­nos­tic meth­ods for patients with Alzheimer’s dis­ease and asso­ci­at­ed disorders.

“Lever­ag­ing AI allows us to look at expo­nen­tial­ly more dis­ease rel­e­vant fea­tures, a pow­er­ful approach when applied to a com­plex sys­tem like the human brain,” said co-cor­re­spond­ing author Kurt W. Far­rell, PhD, Assis­tant Pro­fes­sor of Pathol­o­gy, Mol­e­c­u­lar and Cell-Based Med­i­cine, Neu­ro­science, and Arti­fi­cial Intel­li­gence and Human Health, at Icahn Mount Sinai. 

Read: AI is earning trust as much as humans to flag hate speech

“It is crit­i­cal to per­form fur­ther inter­pretabil­i­ty research in the areas of neu­ropathol­o­gy and arti­fi­cial intel­li­gence, so that advances in deep learn­ing can be trans­lat­ed to improve diag­nos­tic and treat­ment approach­es for Alzheimer’s dis­ease and relat­ed dis­or­ders in a safe and effec­tive man­ner,” added W. Farrell.

“Inter­pre­ta­tion analy­sis was able to iden­ti­fy some, but not all, of the sig­nals that the arti­fi­cial intel­li­gence mod­els used to make pre­dic­tions about cog­ni­tive impair­ment. As a result, addi­tion­al chal­lenges remain for deploy­ing and inter­pret­ing these pow­er­ful deep learn­ing mod­els in the neu­ropathol­o­gy domain,” said the study’s lead author, Andrew McKen­zie, MD, PhD, Co-Chief Res­i­dent for Research in the Depart­ment of Psy­chi­a­try at Icahn Mount Sinai. As a result, there are still more dif­fi­cul­ties in apply­ing and under­stand­ing these potent deep learn­ing mod­els in the field of neuropathology.

This new study pro­vides an impor­tant step for­ward in the devel­op­ment of AI mod­els that can be used to pre­dict cog­ni­tion clin­i­cal­ly from tis­sue slides. Future work will involve con­tin­u­ing AI and deep learn­ing work on more stan­dard histopatho­log­ic fea­tures as well as oth­er brain regions rel­e­vant to cognition.

Also involved in this study were researchers from the Uni­ver­si­ty of Texas Health Sci­ence Cen­ter in San Anto­nio, Texas; New­cas­tle Uni­ver­si­ty in Tyne, Eng­land; Boston Uni­ver­si­ty School of Med­i­cine in Boston; and UT South­west­ern Med­ical Cen­ter in Dallas.

One of the largest aca­d­e­m­ic med­ical sys­tems in the New York met­ro­pol­i­tan area is Mount Sinai Health Sys­tem, which employs over 43,000 peo­ple across eight hos­pi­tals, over 400 out­pa­tient ser­vices, around 300 labs, a school of nurs­ing, a lead­ing med­ical school, and grad­u­ate education.

For more infor­ma­tion, vis­it the cit­ed source.

Leave a Reply

Your email address will not be published.

Join our NewsletterDaily Glimple of Future

Our blog, "Daily Glimpse of Future", strives to make the future much clearer than it is today. Join our newsletter for free now.