A computer system that can detect dementia at an early stage with 85 percent accuracy by analyzing a person’s voice and speech patterns has been developed by IBM’s research lab in Haifa.
Diagnosis is through a series of tests conducted using a smartphone or tablet computer and takes only a few minutes.
Development of the system began in 2012 by IBM researchers, with the participation of 10 academic, health, and information technology organizations. The technology is partially based on IBM’s Watson cognitive computer system, which can recognize human speech. The system also uses dedicated algorithms developed by IBM Israel.
Dementia is a general name for numerous diseases characterized by a gradual reduction in cognitive functioning. The most common such disease is Alzheimer’s, which is responsible for some 60 percent of dementia cases. Some 100,000 Israelis have dementia, and its prevalence increases with lengthening life expectancies. Most dementia sufferers are aged 65 and over and the rate of morbidity rises with age; some 45 percent of people aged 85 and over suffer from dementia.
Dementia is not curable but because it progresses slowly, early detection allows treatments to begin that reduce the symptoms and enable the patients to function better for a longer period of time, as well as live independently longer before they require full-time nursing care. Dementia today is generally detected by psychological and psychiatric tests along with clinical examinations (aimed primarily at eliminating the possibility of metabolic or infectious diseases that could cause dementia). The diagnostic process also includes gathering information from the patient and his or her family, and documenting complaints about forgetfulness and decreased cognitive function.
“The idea is to help identify dementia early,” explains Dr. Aharon Satt, an expert in detecting speech pathologies at IBM’s research division. “We know there is no cure for the disease yet but there is significance in early detection. If we catch it at an early stage, we can treat its symptoms, whether by drug therapy that improves the function of neurotransmitters in the brain or other methods, or by starting psychological and behavioral therapy for the patient and those around him.”
The IBM researchers used the standard diagnostic protocol and created an automated version based on the person’s speech, developing an application that works on a smartphone and a tablet.
“We started with a large set of tests and in the end we reached a point in which we can do a simple and smart examination using a mobile phone or a tablet and get a good indication within a few minutes,” says Satt. The test takes seven minutes, during which the subjects are asked to perform various cognitive tasks, like counting backwards, describing a picture shown to them, identifying words that start with a certain letter, memory tests and more. The examination is recorded and analyzed using a series of algorithms developed by the researchers.
“The more advanced the dementia, the chance of the subject making errors increases, but we don’t look at the mistakes but at the speech characteristics related to the pace of speech, like continuity, hesitation, and fluency,” Satt says. After examining statistical profiles of pauses between words, “We have built a statistical profile that is accurate and sensitive enough to distinguish between a healthy person, a person suffering from MCI [mild cognitive impairment], and patients who are in the first stages of the disease.”
The researchers got assistance from leading European centers for the treatment of dementia, including the Nice Resources and Research Center in Nice, France, and the Greek Association of Alzheimer’s Disease and Related Disorders in Salonika, as well as institutions in Sweden, Ireland, and elsewhere. The researchers collected and analyzed 150 audio recordings of examinations performed by people ranging in age from below 70 to 90.
The subjects had previously undergone comprehensive testing in the usual fashion, which established that a third of them were healthy, a third had MCI, and a third had early-stage dementia. The IBM study, which was published recently in the periodical Dementia & Alzheimer’s, showed that the system diagnosed the participants with 85 percent accuracy.
Though it isn’t clear when the system will be marketable, it is liable to revolutionize the diagnosis of dementia, because it can detect it early and because of its ease of use and low cost compared to today’s diagnostic procedures. The researchers hope that in the future, the system will also be able to aid in diagnosing Parkinson’s, depression, anxiety and psychiatric illnesses that manifest themselves in speech.