Researchers at Mount Sinai recently published research that could improve neuroscientists’ understanding of the progression of Alzheimer’s and other forms of cognitive decline.
The study, published Tuesday in Acta Neuropathologica Communications, uncovers a potential new target that can provide more clues as to why a person goes into cognitive decline. It proved that white matter may be a more important region to look at than was previously thought.
To produce this finding, the research team used a machine learning algorithm, as opposed to traditional biomarkers like amyloid plaques. The algorithm used multiple-instance learning (MIL), a type of supervised machine learning, Kurt Farrell, one of the study authors, said in an interview.
If they were using a non-MIL algorithm, the researchers would have to go in and hand annotate whenever they found a particular brain pathology in the slides, he said. But with a MIL approach, the research team was able to assign labels to the slides — in this case, it was the presence or absence of cognitive impairment. The algorithm then figured out what in each slide differentiated the two labels it had been assigned. This is a more unbiased approach than non-MIL algorithms, Farrell said.
The algorithm was trained by examining structural and cellular features of postmortem human brain tissue samples that had been donated by more than 700 people who had experienced cognitive decline before their death.
The samples were taken from the medial temporal lobe and frontal cortex — these regions of the brain are the most relevant to cognitive impairment, according to Farrell. When collecting these samples from biorepositories across the country, he said his team focused on a group of individuals that had aged and had hyperphosphorylated tau and a lack of beta amyloid, two hallmarks of Alzheimer’s disease. The former occurs when the multiple phosphorylation sites on tau protein cells become fully saturated, and beta amyloid is the chief component of the amyloid plaques found in the brains of people who have Alzheimer’s.
After the algorithm was trained and ran on a dataset, the team came across an interesting finding: the algorithm produced heat maps showing that the top reason it made the decision whether or not someone was cognitively impaired during their life had to do with the white matter in the sample.
This was a “rather striking” finding because the research team’s working hypothesis going into the study was that the algorithm would likely focus on the gray matter, Farrell said. He pointed out that gray matter is often examined in neuropathologic research for Alzheimer’s, whereas neurologists haven’t really looked to white matter for clues yet.
“To distill it down, the algorithm felt that a region of the brain that we don’t often look at was actually more important than the region that we do look at,” Farrell said.
Alzheimer’s and other dementia-causing conditions are multifactorial diseases, so having white matter as a new research target opens the door for neuroscientists to potentially find new involvement in the brain that could lead to cognitive decline. That makes the study’s finding an “exciting piece of knowledge,” according to Farrell.
He noted that this research is early, and there are many more studies that must be conducted to fully understand white matter’s role in cognitive decline.
“If we carefully examine the slides from those who have graciously donated their brain to these postmortem tissue samples, we can provide some clues as to what was happening in their brain during their life,” Farrell said. “But this is only one time point. We’re hoping that the research we’re conducting can provide clues to others who are doing neuroimaging and research on living subjects and help them look at new targets.”
Photo: SIphotography, Getty Images