Liang is a professor of biomedical engineering in the School of Biomedical Engineering, Science & Health Systems.
Agbavor, a doctoral student, collaborated on the research.
A team from Drexel’s School of Biomedical Engineering, Science and Health Systems has used a popular chatbot program to identify clues from spontaneous speech that predict early stages of dementia, with 80% accuracy.
The team’s work, reported in the journal PLOS Digital Health, is the latest in a series of efforts to show the effectiveness of natural language processing programs in spotting speech delays and other verbal tics that are indicative of neurodegenerative disorders.
The researchers tested their theory by training OpenAI’s GPT-3 with recorded speech transcripts from people exhibiting varying stages of Alzheimer’s decline. From that, GPT-3 generated a characteristic profile of Alzheimer’s speech. That was then used to retrain the program — turning it into an Alzheimer’s disease screening machine.
The process was demonstrated to reliably predict those who have Alzheimer’s disease and those who don’t.
The team also found that GPT-3 was 20% more accurate in predicting the severity of disease than an alternative analysis based solely on the acoustic features of the recordings, such as pauses, voice strength and slurring.
The team’s findings could speed Alzheimer’s diagnosis, currently a lengthy process that involves a medical history review and a host of physical and neurological tests. Spotting the disease early would give patients more options for therapeutics and support.
To build on these promising results, the researchers are planning to develop a pre-screening tool that could be used at home or in a doctor’s office.