AI and Writing May Dramatically Increase Early Dx of Alzheimer’s
Telltale strokes, loops, and tremors plus repeated words are now offering new AI clues to diagnose Alzheimer’s.
Sending a handwritten letter is becoming such an anomaly. It’s disappearing. My mom is the only one who still writes me letters. And there’s something visceral about opening a letter — I see her on the page. I see her in her handwriting. — Steve Carell
Medical archives, similar to family photo albums, can provide valuable clues to medical illnesses long before their symptoms present. They are gold for mining when researchers begin to explore the rich trove waiting for them, and the results may provide advances in medicine.
One such invaluable cache is the data collected for The Framingham Heart Study (FHS), which came into existence in 1948. “Researchers recruited 5,209 men and women between the ages of 30 and 62 from the town of Framingham, Massachusetts.”
Limited in scope initially to collect data on cardiovascular disease factors, its data bank has spread into a 14K three-generational study of more than heart disease. One unexpected area is a potential key to unlocking the mysteries of early diagnosing Alzheimer’s.
“…the original goal of the Framingham Heart Study (FHS) was to identify common factors or characteristics that contribute to cardiovascular disease. Over the years, the FHS has become a successful, multigenerational study that analyzes family patterns of cardiovascular and other diseases, while gathering more genetic information from the two generations that followed the original study participants…”
AI Joins the Hunt
Previous efforts to assess neurologic disorders have depended, to a limited extent, on handwriting or geometric design drawing samples. A patient was asked to copy a design or write a simple sentence of their choosing, and then a visual assessment was made. Results were dependent on the individual doing the assessment and prone to error as a result.
Technology has provided a new approach to one specific paper-and-pencil test, the Bender-Gestalt (BGT), which would use automated image analysis and computerized scoring. “A comparison of the scores reported by the automated system with those assigned by the psychologists not only reveals the effectiveness of the proposed system but also reflects the huge research potential this area possesses.”
The subtle design differences in patient replication can more easily be detected by an automated program than the naked eye as this study appears to indicate.
Owing to current diagnostic testing challenges for neurologic disorders such as Alzheimer’s, the National Institute on Aging‐Alzheimer’s Association workgroup has recommended changes in criteria and methods. But quick assessments still need to be performed before more rigorous testing with the expensive armamentarium of machinery like MRI, CT, or PET scans.
The current emphasis is directed at predicting rather than diagnosing Alzheimer’s, hoping that it can be prevented or at least forestalled. It is here that both language and writing samples early in adult life may be useful indicators. The existence of a database created for The Framingham study plays a significant role because of the size of the sample and the length of the study.
A New Use for Existing Data
A research team utilized the Framingham Study base and “…applied state of the art analysis procedures to samples from 270 participants finding that dementia onset before age 85 could be identified with 70–75% accuracy. This is well above chance performance, and slightly below the best performance seen in studies comparing current dementia patients with controls.”
Their results received support from an extensive literature review. “The systematic literature review supported the argument that language and speech could successfully be used to detect dementia automatically.”
Years prior to the appearance of symptoms, writing and language production were found to be useful predictors of illness.
A joint research effort by IBM and Pfizer revealed the efficacy of utilizing AI and speech/writing samples early in adult life. The outstanding linguistic features of those who would develop Alzheimer’s in the future included:
- misspelled or inappropriately capitalized words
- telegraphic language meaning simple two-word sentences
- simple grammatical structure with missing subjects and words such as “the,” “is,” and “are”
Prior studies have established the presence of writing and grammar difficulties in those with Alzheimer’s.
“Performance analysis of the dictation written spelling task for ATD (Alzheimer’s type dementia) patients shows a significant impairment compared to control subjects both for words (regular or irregular) and nonwords, and for error types, which confirms that writing is commonly impaired in ATD.”
The increasing utility of AI in medicine and diagnosis is apparent. However, the search for cures and effective treatment remains. Anyone agreeing to be tested for a future diagnosis of a debilitating disease will still have to make a difficult decision in the interim.