(Arthur Toga, UCLA/NIH.gov)
23 August 2017. Neurologists and computer scientists developed techniques for analyzing images to determine the chance of someone progressing from mild cognitive impairment to Alzheimer’s disease. The team from McGill University in Montreal, Quebec, Canada describe these methods in the 10 July issue of the journal Neurobiology of Aging.
The researchers led by neurologists Pedro Rosa-Neto and Serge Gauthier are seeking to determine with greater certainty those individuals at the greatest risk of developing Alzheimer’s disease, even in early stages of mild cognitive impairment. Alzheimer’s disease is a progressive neurodegenerative condition affecting growing numbers of older people worldwide. People with Alzheimer’s disease often have deposits of abnormal substances in spaces between brain cells, known as amyloid-beta proteins, as well as misfolded tangles of proteins inside brain cells known as tau.
With current tools, neurologists still cannot predict with certainty which individuals are likely to progress into Alzheimer’s disease, even with positron emission tomography or PET scans that highlight accumulations of amyloid-beta proteins. The problem is particularly acute when recruiting participants in clinical trials of treatments for Alzheimer’s disease, which have limited time frames and need to carefully enroll the optimum candidates for new therapies. The goal of the team’s project, therefore was to devise a way of determining the chance of an individual developing Alzheimer’s disease in the next 24 months based on a PET scan of that person’s brain.
Rosa-Neto, Gauthier, and colleagues at McGill took advantage of a long-term research project known as the Alzheimer’s Disease Neuroimaging Initiative to craft a solution. This initiative recruits individuals over the age of 55 worldwide, both healthy and with mild memory problems, as well as people diagnosed with Alzheimer’s disease. These individuals are then tracked over a number of years, reporting changes in cognition, function, brain structure, and biomarkers.
The team sampled cases from the Alzheimer’s Disease Neuroimaging Initiative to construct an algorithm for predicting the onset of Alzheimer’s disease. The researchers selected 273 participants in the database, covering both PET scan images and detailed biomarker data as raw material. Computer scientists in the lab then used this material to train a predictive algorithm with machine-learning, linking characteristics of the individuals’ PET scans with biomarker indicators and decline in cognition and function associated with the development of Alzheimer’s disease.
In tests of the algorithm against real cases, the researchers found the algorithm could analyze a patient’s PET scan and accurately predict the onset of Alzheimer’s with 84 percent accuracy within the next 24 months, with other characteristics of the disease largely matching up as well. The team says its algorithm outperforms other algorithms when using the same biomarkers, as well as earlier studies analyzing biomarker measurements.
While the algorithm is expected to immediately aid in recruiting participants for clinical trials, the researchers believe it can have benefits for patients as well, to help plan for eventual progression of the disease. “This is an example how big data and open science brings tangible benefits to patient care,” notes Rosa-Neto in a university statement. Clinical use of the algorithm, however, must first meet regulatory requirements in Canada and elsewhere.
More from Science & Enterprise:
- Computer-Designed Antibodies Target Alzheimer’s Proteins
- LEDs Shown to Reduce Alzheimer’s Brain Plaques
- Trial to Test Radio Waves to Treat Alzheimer’s
- Safety, Effects Shown in Drug Clearing Alzheimer’s Peptides
- Trial Shows Antibody Slows Alzheimer’s Cognitive Decline
* * *
RSS - Posts
You must be logged in to post a comment.