AI Machine Learning Predicts Alzheimer’s Risk

Machine Learning

The most common cause of dementia worldwide is Alzheimer’s disease (AD), a neurodegenerative disease with no known cure.New research published in scientific report uses artificial intelligence (AI) machine learning (ML) and electronic health record (EHR) data to identify key predictors of Alzheimer’s disease, including individual I discovered that genetics trump age.

Ad followed by article

“Machine learning (ML) techniques are an attractive and effective alternative to traditional statistical regression models, especially when there are a large number of features and predictors.” Xiaoyi Raymond Gao, Ohio State University School of Medicine, Ohio State University researchers Marion Chiariglione, Ke Qin, and Douglas Scharre. University of Miami researchers Karen Nuytemans and Eden Martin. Yiju Lee at Duke University.

According to the World Health Organization (WHO), Alzheimer’s disease accounts for an estimated 60-70% of the more than 55 million people with dementia worldwide, and it affects women particularly hard.

According to the Alzheimer’s Association, 6.7 million people over the age of 65 currently have Alzheimer’s disease in the United States, nearly two-thirds of whom are women, and by 2050 that number is projected to reach 12.7 million. increase significantly.

Alzheimer’s disease was first identified in 1906 by the German psychiatrist and neurologist Alois Alzheimer. He found abnormal masses and tangled bundles of fibers in the brain tissue of a female patient, Auguste Déter, who died at the age of 51. Alois Alzheimer was being treated for amnesia, irrational behavior and communication problems at a psychiatric hospital in Frankfurt. Today, these abnormal clumps are known as amyloid plaques, and the tangled bundles of fibers are known as neurofibrillaries or tau tangles.

Ad followed by article

In addition to memory problems, other AD symptoms include problems thinking, reasoning, making decisions, judging, and performing daily tasks, according to the Mayo Clinic. Alzheimer’s disease is characterized by personality and behavioral changes with symptoms including depression, delusions, altered sleep habits, loss of inhibition, mood swings, anger, aggression, loss of interest in activities, social withdrawal, and wandering. can cause There is no cure, but the progression of AD symptoms can be slowed with medication from the Mayo Clinic.

Alzheimer’s brain changes can occur more than a decade before symptoms appear, according to the National Institute on Aging. Early detection of the disease not only enables AD patients and their caregivers to plan future care services, but also provides opportunities for treatment of symptoms that may help improve quality of life. increase.

In this new study, the researchers used a popular machine learning library called eXtreme Gradient Boosting (XGBoost) and Shapley Additive exPlanations (SHAP), a state-of-the-art algorithm for explainability in AI machine learning. We aimed to create an explainable AI model. Reverse engineer the output of a prediction algorithm based on the game theoretical optimal Shapley value. SHAP computes the contribution of each feature to the prediction, so it is a useful tool for visualizing the output. Researchers used over 11,000 features and predictors.

Ad followed by article

“The combination of XGBoost and SHAP can be used as an explainable ML model. is provided, ”the scientist wrote.

Researchers generated the Alzheimer’s disease polygenic risk score (PRS) from the Alzheimer’s Disease Genetics Consortium database and the age at onset of AD (AAO) using the UK Biobank database.

“Although the apolipoprotein E gene (APOE) is the best-known genetic risk factor for AD3,12, genome-wide association studies (GWAS) have identified more than 40 loci in AD to date. ’” shared the researchers. “Recently, the polygenic risk score (PRS) has been proposed to aggregate genetic influences across the genome, from small to large, into a single measure of risk for each individual.”

An AI Machine Learning Model for Predicting Alzheimer’s Disease Was Developed Using International Classification of Diseases Tenth Revision (ICD-10) Codes from Electronic Health Records (EHRs) and Genetic Data from Large Biorepositories .

“To the best of our knowledge, this is a large cohort study using a state-of-the-art explainable ML framework, using genetic and non-genetic information and the ICD-10 code from the EHR to identify AD. This is the first report to develop a predictive model,” the researchers wrote.

Ad followed by article

Researchers found age, income, and polygenic risk score to be the top AD risk factors that improve predictive accuracy.

Other important risk factors include family history of Alzheimer’s disease/dementia, hearing impairment, diabetes, and blood pressure (increased systolic and decreased diastolic). Interestingly, they found that being underweight, but not obese, increases the risk of Alzheimer’s disease and may be a useful preclinical biomarker.

A key finding is that electronic medical record data can provide important data for predicting Alzheimer’s disease, and that the AI ​​model provided the top 20 features for both the 40+ and 65+ age groups. . Scientists point out that feature importance does not indicate causality.

The AI ​​machine learning model ranked first among all features in the age group of 40 years and older, and the genetic influence reflected in the polygenic risk score was compared to that of the age group of 65 years and older to predict Alzheimer’s disease. made clear that it becomes more important than the individual’s age.

Copyright © 2023 Kami Rosso All rights reserved.

Source link

Leave a Reply

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