Detecting Alzheimer’s disease sooner means preparations, and perhaps even preventative measures, can be put in place, say scientists, who predict that shortly, artificial intelligence (AI) models could provide an early warning in people destined to develop symptoms years before they appear.
A team from the University of California-San Francisco (UCSF) and Stanford University applied machine learning methods to more than 5m health records, training the AI to spot patterns that connect Alzheimer’s to other conditions.
While the resulting system isn’t perfect, when tested against records for people who developed Alzheimer’s later, the AI accurately predicted its development 72% of the time – up to seven years prior, in some cases.
The system’s predictive power stems from its ability to combine analyses of several different risk types to calculate the likelihood of Alzheimer’s developing.
“This is a first step towards using AI on routine clinical data, not only to identify risk as early as possible, but also to understand the biology behind it,” said bioengineer Alice Tang, from UCSF.
The model detected a number of conditions that could be used to calculate Alzheimer’s risk, including high blood pressure, high cholesterol, vitamin D deficiency, and depression. Erectile dysfunction and an enlarged prostate were also significant factors in men, with osteoporosis significant for women.
That’s not to say people with these health issues will develop dementia, but the AI analysis weighs each as predictors worth looking at. It’s hoped that the same kind of machine learning approach might one day be able to identify risk factors for other hard-to-diagnose diseases.
“The combination of diseases is what allows our model to predict Alzheimer’s onset,” says Tang. “Our finding that osteoporosis is one predictive factor for females highlights the biological interplay between bone health and dementia risk.”
The researchers also investigated the biology behind some of the identified links. Osteoporosis, Alzheimer’s in women, and a variant in the gene MS4A6A, were found to be connected, providing new opportunities to study the disorder’s development.
“This is a great example of how we can leverage patient data with machine learning to predict which patients are more likely to develop Alzheimer’s, and also to understand the reasons why that is so,” said Marina Sirota, a computational health scientist at UCSF.
The research was published in Nature Aging.
Study details
Leveraging electronic health records and knowledge networks for Alzheimer’s disease prediction and sex-specific biological insights
Alice Tang, Katherine Rankin, Marina Sirota et al.
Published in Nature Aging on 21 February 2024
Abstract
Identification of Alzheimer’s disease (AD) onset risk can facilitate interventions before irreversible disease progression. We demonstrate that electronic health records from the University of California, San Francisco, followed by knowledge networks (for example, SPOKE) allow for (1) prediction of AD onset and (2) prioritisation of biological hypotheses, and (3) contextualisation of sex dimorphism. We trained random forest models and predicted AD onset on a cohort of 749 individuals with AD and 250,545 controls with a mean area under the receiver operating characteristic of 0.72 (7 years prior) to 0.81 (1 day prior). We further harnessed matched cohort models to identify conditions with predictive power before AD onset. Knowledge networks highlight shared genes between multiple top predictors and AD (for example, APOE, ACTB, IL6 and INS). Genetic colocalisation analysis supports AD association with hyperlipidaemia at the APOE locus, as well as a stronger female AD association with osteoporosis at a locus near MS4A6A. We therefore show how clinical data can be utilised for early AD prediction and identification of personalised biological hypotheses.
See more from MedicalBrief archives:
Re-imagining medicine’s future with the help of digital technology
Spotting brain changes 10 years before Alzheimer’s symptoms appear
Smartwatches may predict early Parkinson’s – UK study
Neural-network model can accurately predict depression