In Spain, Alzheimer’s disease affects approximately 800,000 people, and around 40,000 new cases are diagnosed per year, according to the Spanish Society of Neurology (SEN). Taking into account that 80% of the cases that are mild are undiagnosed, it is important to develop new methods that contribute to an early detection of the disease. In this sense, IBM Research has carried out together with Pfizer a model based on artificial intelligence (AI) capable of predicting with 71% accuracy the appearance of Alzheimer’s in healthy people.
This research, published in The Lancet eClinical Medicine , offers important advances in finding different pathways to predict Alzheimer’s. The vast majority of existing research on the prediction of this disease has focused on people who were beginning to show signs of cognitive decline, or those with risk factors such as family history or genetics. However, Alzheimer’s disease can affect a wide spectrum of people, including those without a family history or other risk factors. That is why the IBM Research study is one of the first to use AI to predict outcomes in healthy people with no other risk factors at play.
Linguistic markers
The AI model uses small non-invasive samples of the patient’s language obtained through cognitive tests. Using linguistic markers, the artificial intelligence model produces valuable predictive results, with a higher accuracy rate than predictions on a clinical scale (59%) based on other available biomedical data.
This model has been trained using long-term longitudinal data from the Framingham Heart Study , a large multigenerational study that began in 1948 and has prompted thousands of health studies. Due to the nature of these data, the researchers were able to verify the predictions of their model with the actual results. For example, if the AI model analyzed a sample of the language of a 65-year-old participant and predicted that they would develop the disease at age 85, the researchers could check the registries to determine if that diagnosis was made.
Extended data set
One of IBM’s goals with this research is to be able to train the model using extended data sets. That would include data that could be available to health researchers in a short time and would reflect a broader geographic, socioeconomic and racial diversity of data. As a result, the algorithms of the IBM researchers could serve as a potential clinical asset for healthcare professionals when assessing a holistic view of an individual’s health and risk factors.
This work is part of a larger platform that IBM Research is building to better understand neurological health and chronic diseases through biomarkers and signals in speech and language. Research on Alzheimer’s will be a significant pillar on this platform and will advance this work, which is now moving into prospective studies. IBM Research recently announced an initiative related to the National Institutes of Health to analyze speech and discover risk factors for schizophrenia and psychosis.
Artificial intelligence
IBM Research continues to explore ways that artificial intelligence can be applied to predict a person’s risk of Alzheimer’s disease, which as of now has no effective cure or prevention.
Up to 5.5 million Americans are estimated to have Alzheimer’s disease, and recent studies suggest that it may be the third leading cause of death in the US after heart disease and cancer.
