IBM Leverages Machine Learning for Blood-Based Alzheimer's Screening with 77% Accuracy

VCBeat (WeChat Official Account: vcbeat) has learned that on March 11, 2019, IBM’s Australian research team announced the development of an alternative to cerebrospinal fluid testing for Alzheimer’s disease. This method employs machine learning-based blood tests to determine the level of amyloid-beta plaques in cerebrospinal fluid, achieving an accuracy rate of up to 77%.

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder with an insidious onset. Clinical manifestations include memory impairment, aphasia, apraxia, agnosia, visuospatial deficits, executive dysfunction, and changes in personality and behavior. Onset before age 65 is termed presenile dementia, while onset after age 65 is termed senile dementia.

In traditional clinical practice, protein detection requires invasive procedures performed under anesthesia; this complex process has precluded large-scale early detection of Alzheimer’s disease.

IBM’s Australian research team stated that using machine learning to study four proteins in the blood of individuals who may carry a genetic risk for Alzheimer’s disease would help identify their risk of developing the condition and improve the conduct of clinical trials for candidate drugs.

An IBM research team in Australia downloaded data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), a long-term imaging study focused on older adults, to investigate whether MRI and PET images could be combined with other measures for predicting early-stage Alzheimer’s disease using machine learning. They examined 566 participants in the ADNI study, including 182 carriers of the APOE ε4 allele. APOE ε4 is a genetic variant that increases the risk of late-onset Alzheimer’s disease.

Researchers can define Alzheimer’s disease by the relationship between amyloid-beta plaques and tau tangles, protein aggregates that share characteristics with prions. The strange “infectious proteins” discovered by Prusiner in the 1980s may be the true culprit behind Alzheimer’s disease. As prions self-replicate and accumulate, they also trigger various neurodegenerative disorders, such as Creutzfeldt-Jakob disease.

Current research has demonstrated that tau protein and amyloid-beta self-assemble into aggregates. Similar to the role of proteins in prion diseases, these aggregates originate from a single abnormal molecule and subsequently propagate throughout the human brain.

“Although blood analysis methods are applicable to people of all age groups, currently, protein level measurement techniques can achieve the best predictive performance,” wrote the IBM Australia research team in their report.

“As life expectancy continues to rise, neurodegenerative diseases such as Parkinson’s disease, Alzheimer’s disease, and Huntington’s disease are affecting millions of people worldwide,” said Smith Gaudy, a member of the IBM Australia research team. “Although there is currently no reliable cure for these conditions, we have made significant progress in early prevention and alleviating symptoms.”

(Compiled by Hu Xuan)