Machine-Learning Approach May Diagnose Alzheimer's Disease
A selective assessment method may lead to earlier diagnosis and treatment of Alzheimer’s disease.
A novel machine-learning strategy may be better suited to diagnose Alzheimer’s disease prior to symptom onset compared with current methods, according to a study published by Scientific Reports.
Current estimates indicate that more than 5 million Americans have Alzheimer’s disease, with the numbers continuously increasing due to an aging population. While there is no cure for the progressive condition, several drugs can slow or prevent symptom worsening for up to 5 years, which allows individuals to remain independent.
The key to slowing or preventing disease progression is early diagnosis and treatment; however, current methods may not be able to detect the condition before symptom onset.
The novel machine-learning computer program approach includes numerous Alzheimer’s disease indicators, including mild cognitive impairment, according to the study.
“Many papers compare the healthy to those with the disease, but there’s a continuum,” said Anant Madabhushi, PhD. “We deliberately included mild cognitive impairment, which can be a precursor to Alzheimer’s, but not always.”
The novel Cascaded Multi-view Canonical Correlation (CaMCCo) algorithm was examined using data from 149 patients included in the Alzheimer’s Disease Neuroimaging Initiative.
The algorithm integrates several factors to determine Alzheimer’s disease, including MRI scans, features of the hippocampus, glucose metabolism rates in the brain, proteomics, genomics, mild cognitive impairment, and other parameters, according to the study.
Previously, the authors discovered that integrating dissimilar information can accurately identify cancers. This is the first time they have done so for Alzheimer’s disease.
“The algorithm assumes each parameter provides a different view of the disease, as if each were a different set of colored spectacles,” Dr Madabhushi said.
The technique evaluates the factors in a 2-stage cascade: first, CaMCCo selects the variables that indicate whether a patient is healthy or unhealthy; second, it selects the variables that determine whether a patient has mild cognitive impairment and which patients have Alzheimer’s disease, according to the study.
“The remaining views are combined to give the best picture,” Dr Madabhushi said.
When looking at Alzheimer’s disease, CaMCCo was observed to outperform each individual factor and other methods that look at a combination of the factors, according to the study. The novel method also better predicted which patients had mild cognitive impairment compared with methods that also evaluated multiple factors.
The authors plan to further validate CaMCCo using additional data to ensure that it is highly accurate. Additionally, they plan to run data for neurologists as they compile patients’ tests.
If the novel test is validated, the authors expect to conduct clinical trials to determine its accuracy in predicting early Alzheimer’s disease, the study concluded.