Pfizer and IBM's AI Model Predicts Alzheimer's Onset with 71% Accuracy Using Linguistic Biomarkers

Alzheimer's Disease (AD), also known as senile dementia, is a neurodegenerative disease with an insidious onset and progressive development; the average survival period for patients is 5.5 years. Although this disease is more common in elderly people over the age of 70, symptoms and responses actually begin to appear as early as middle age. Currently, there is no cure for Alzheimer's disease, thereforeEarly detection and treatment are key to managing this disease.

Recently, U.S. pharmaceutical giantPfizerThe Company and IT GiantsIBMThe company jointly published an article titled “Linguistic markers predict onset of Alzheimer's disease” in The Lancet EClinicalMedicine, offering an intelligent option for the early detection of Alzheimer’s disease. The article states thatAn artificial intelligence (AI) model jointly developed by two companies can help predict the eventual onset of Alzheimer’s disease with 71% accuracy based on language samples, potentially transforming the current landscape of Alzheimer’s diagnosis, treatment, and drug development.

doi.org/10.1016/j.eclinm.2020.100583

Previously, IBM had tested the reliability of using artificial intelligence to identify proteins for predicting β-amyloid concentrations, and other studies were also underway to design methods capable of predicting the progression of Alzheimer’s disease. The distinction of this study by IBM and Pfizer lies in its use of data collected before participants first exhibited symptoms of cognitive impairment, as well as its assessment of dementia risk in the general population rather than in high-risk groups.

Specifically, this study aimed to assess the extent to which language proficiency at a single time point can serve as a prognostic indicator for future conversion to Alzheimer’s disease (AD). The researchers utilized data from the US Framingham Heart Study, initiated in 1948, analyzing 703 samples from 270 participants. The linguistic analysis incorporated 87 variables, including spelling errors, verbosity, lexical diversity, and repetition, while also accounting for assessments of age, sex, education level, visuospatial ability, and executive reasoning.

Methods for Case and Control Selection and Prediction Model Setup

Researchers analyzed speech samples from 80 participants before they exhibited signs of cognitive decline. The final results showed that half of these participants developed Alzheimer’s disease (AD) before the age of 85. The linguistic variables most strongly associated with predicting AD onset were telegraphic speech (characterized by grammatical errors), repetitiveness, and spelling errors. The mean time from normal cognition to a diagnosis of mild AD was 7.59 years, and the mean time from normal cognition to the onset of cognitive impairment was 3.93 years.

Researchers believe that this test may not necessarily replace current clinical standards, but it can be used as an early-step procedure to recommend brain scans. Additionally,This study may play a pivotal role in facilitating the development of novel Alzheimer’s disease (AD) therapies, thereby assisting drug developers in designing more robust clinical trials for Alzheimer’s disease.。

“A review of past failures in Alzheimer’s drug development reveals that most enrolled patients were in the midst of disease progression. Digital biomarkers capable of quantitatively assessing an individual’s disease progression will undoubtedly provide stronger support for Alzheimer’s clinical trials,” said Ajay Royyuru, IBM Fellow and Vice President of Healthcare and Life Sciences Research at IBM.

References:

[1] Linguistic markers predict onset of Alzheimer's disease.

[2] Could a language test diagnose Alzheimer's? IBM is laying the groundwork.

Responsible Editor: Liuli

Disclaimer: This article is reposted from Yaozhi.com. The copyright belongs to the original author. It is intended solely for the purpose of information dissemination and does not represent the views of this platform. If there are any issues concerning the content, copyright, or other matters, please contact us by leaving a message on our website. We will remove the content promptly upon verification.