Analyzing language predicts Alzheimer’s years in advance
There are over 500,000 Canadians living with dementia today, with 25,000 newly diagnosed cases each year. These numbers come from the Alzheimer’s society of Canada, which also reports that 56% of Canadians are concerned about developing dementia. Many people report being aware of subtle changes in memory, focus or behaviour over time. While in some cases these are simply age-related and no cause for concern, in others they indicate Mild Cognitive Impairment (MCI), which can sometimes represent early Alzheimer’s Dementia (AD).
In a recent study, researchers at IBM and Pfizer created a computer model to analyze the written responses of participants in the Framingham Heart Study who were healthy at the time. The Framingham study participants are a historic cohort, as they have been closely monitored since 1948, and they and their offspring have undergone regular testing ever since. This model scored their responses in 87 different ways, and researchers compared the model’s predictions with actual cases of AD diagnosed before 85 years old (EClinicalMedicine doc.org/10.1016/j.eclinm.2020.100583).
For people who take a proactive approach to restoring and maintaining their health and wellness, brain function should obviously be a priority. On many levels, a better brain means a better life. This can be achieved through mindful nutrition and healthy food choices, regular exercise, fresh air, sunshine, time in nature, healthy relationships and social life, and mindfully managing our thoughts and emotions in our daily lives. Many people have turned to natural health products, manual therapies, acupuncture or other kinds of integrative treatment.
While there are no silver bullets that can reverse AD, there is a critical need for earlier diagnosis. Researchers have studied blood tests, MRI imaging, cerebrospinal fluid and neuropsychological test scores to try and find early markers of the disease. In a recent study, researchers combined many of these test results to create a score that predicted which MCI patients would go on to develop AD over the next two years (PLoS ONE doi.org/10.1371/journal.pone.0021896).
This new study took a radically different approach, using computation to analyze language, and while the results were not perfect, they were impressive. Future onset of AD was associated with three specific variables called telegraphic speech, repetitiveness and misspellings, allowing the model to predict AD with about 70% accuracy.
Language skills are a marker of what researchers call ‘cognitive reserve’, which is related to IQ, education, occupation and other factors. Cognitive reserve allows some people to be more resilient to brain pathology. While there has not been enough research in this field, preliminary studies suggest that your daily activities can influence your cognitive reserve and your brain health. Machine learning is advancing rapidly, and one day soon you may be able to predict the future health of your brain with a simple quiz.