Machine learning nuances the stages of Alzheimer’s — ScienceDaily

A Cornell-led collaboration used machine learning to establish the most accurate means and timelines to anticipate the progression of Alzheimer’s disease in people who are cognitively normal or experiencing mild cognitive impairment.

The modeling showed that predicting future decline to dementia is easier and more accurate for individuals with mild cognitive impairment than for cognitively normal or asymptomatic individuals. At the same time, the researchers found that the predictions for cognitively normal subjects are less accurate for longer time horizons, but the opposite is true for individuals with mild cognitive impairment.

The modeling also showed that magnetic resonance imaging (MRI) is a useful prognostic tool for people at both stages, while tools that track molecular biomarkers, such as positron emission tomography (PET) scans, are more useful for people with mild cognitive impairment.

The team’s paper, “Machine Learning Based Multi-Modal Prediction of Future Decline Toward Alzheimer’s Disease: An Empirical Study,” published Nov. 16 in PLOS ONE. The lead author is Batuhan Karaman, a PhD student in electrical and computer engineering.

Alzheimer’s disease can take years, sometimes decades, for a person to show symptoms. Once diagnosed, some individuals rapidly deteriorate, but others can live for years with mild symptoms, making predicting the rate of progression of the disease challenging.

“If we can say with confidence that someone has dementia, it’s too late. A lot of damage has already been done to the brain, and it’s irreversible damage,” said senior author Mert Sabuncu, an associate professor of electrical and computer engineering in the College of Engineering. and from electrical engineering in radiology at Weill Cornell Medicine.

“We really need to be able to detect Alzheimer’s disease early,” Sabuncu said, “and see who is progressing quickly and who is slower, so we can stratify the different risk groups and target whatever treatment options we have.”

Doctors often focus on a single “time horizon” — usually three or five years — to predict a patient’s progression of Alzheimer’s. The time frame can seem arbitrary, according to Sabuncu, whose lab specializes in the analysis of biomedical data — especially imaging data, with an emphasis on neuroscience and neurology.

Sabuncu and Karaman teamed up with longtime collaborator and co-author Elizabeth Mormino of Stanford University to use neural network machine learning that could analyze five years’ worth of data on individuals who were cognitively normal or had mild cognitive impairment. The data, captured in a study by the Alzheimer’s Disease Neuroimaging Initiative, includes everything from a person’s genetic history to PET and MRI scans.

“What we were really interested in is, can we look at this data and see if someone will make progress over the next few years?” Sabuncu said. “And more importantly, can we make better predictions when we combine all the follow-up data points we have on individual subjects?”

The researchers discovered a number of striking patterns. For example, predicting a person going from asymptomatic to mild symptoms is much easier for a one-year time horizon compared to five years. However, predicting whether someone will progress from mild cognitive impairment to Alzheimer’s dementia is most accurate over a longer timeline, with a sweet spot of about four years.

“This could tell us something about the underlying disease mechanism and how it evolves over time, but that’s something we haven’t explored yet,” Sabuncu said.

With regard to the effectiveness of different types of data, the modeling showed that MRI scans are most informative for asymptomatic cases and are especially useful for predicting whether someone will develop symptoms in the next three years, but less useful for prognosis for people with a mild cognitive impairment. . Once a patient has developed mild cognitive impairment, PET scans, which measure certain molecular markers, such as the proteins amyloid and tau, appear to be more effective.

An advantage of the machine learning approach is that neural networks are flexible enough to function despite missing data, such as patients who may have missed an MRI or PET scan.

In future work, Sabuncu plans to further tweak the modeling so it can process full imaging or genomic data, rather than just summary measurements, to collect more information that will increase predictive accuracy.

The research was supported by the National Institutes of Health National Library of Medicine and National Institute on Aging, and the National Science Foundation.

Story Source:

Material supplied by Cornell University. Originally written by David Nutt, courtesy of the Cornell Chronicle. Note: Content is editable for style and length.

Leave a Reply

Your email address will not be published. Required fields are marked *