An algorithm capable of detecting Alzheimer
Posted on February 2, 2021
All efforts to improve the diagnosis and early care of Alzheimer's are few. The Stevens University of Technology in New Jersey has developed an algorithm that detects the language deficits typical of the early stages of the disease.
According to data from the World Health Organisation, dementia affects about 50 million people worldwide, of which about 60% live in low and middle-income countries. Every year about 10 million new cases are registered and it is estimated that 5% and 8% of people over 60 years of age suffer or will suffer from this health problem. It is estimated that in 2030 there will be 82 million people with dementia and 152 million in 2050. Therefore, research on Alzheimer's and other types of cognitive impairment has not stopped growing in recent years.
With no cure and no simple way to diagnose the disease, scientists are exploring all avenues when it comes to diagnosing Alzheimer's in its early stages. Getting down to business, a team of researchers has focused on subtle differences in patients' language, developing an Artificial Intelligence based tool that screens them for potential disease.
The research focuses on the way some Alzheimer's patients express themselves. People in the early stages of the disease have trouble finding the words they are looking for, using synonyms, going blank, or replacing nouns with pronouns.
The purpose of the project was to create a software-based on Artificial Intelligence that could detect these linguistic differences, using a standard image description task that is currently used in the detection of language for Alzheimer's. Subjects must describe a picture of children stealing cookies from a jar. For the smart tool, the team relied on existing transcripts of more than 1,000 interviews, both from Alzheimer's patients and healthy controls.
The AI algorithm based on machine learning was trained with the individual sentences broken down to which numerical values were assigned so that the system could analyse the structural and thematic relationships between them. The algorithm learned to differentiate over time between sentences spoken by healthy and Alzheimer's patients with an accuracy of more than 95%.
The next steps will be to expand the tool for use in languages other than English, and even enable it to diagnose Alzheimer's disease using other types of text, such as an email or a post on social media. Researchers also see great potential in its use to track the progression of Alzheimer's overtime.
Leave a Comment:
Daily Dose of Positivity