Creating an Algorithm for the Brain: Predicting the Prognosis of Alzheimer’s Disease
Written by: Jon Zhang ‘24
Edited by: Meehir Dixit ‘24
Over 5 million Americans currently suffer from Alzheimer’s Disease (AD), a crippling cognitive condition that inhibits important mental functions and interferes with daily life .
Despite widespread knowledge of its symptoms, little else is understood about AD. Along with having no cure, causes and strategies to slow the disease’s progression remain largely unknown. Tragically, individuals are forced to watch as their loved ones slowly succumb to the deteriorating effects of mental decline, withering away into shadows of their former selves.
Scientists have taken great strides to uncover the mysteries surrounding AD. While the data is not conclusive, researchers hypothesize that AD is caused by an abnormal accumulation of proteins called beta-amyloid and tau around brain cells. Medical procedures like PET scans and spinal taps for cerebrospinal fluid examine the levels of these proteins to make an assessment . However, these operations are both expensive and invasive, creating a need for affordable and less-invasive techniques for AD detection.
A recently-published IBM study begins to simplify this process. Currently, clinical assessments, such as the Mini-Mental State Examination and Wechsler Adult Intelligence Scale, have been used to examine cognitive functions such as attention, concentration, and word retrieval to indicate cognitive decline. While current exam variations contain multiple items that can take up to 90 minutes, the IBM study notably includes just one question, taken from the Boston Aphasia Diagnostic Examination .
This greatly-simplified “exam” asked a cohort of 80 participants to observe and describe a situation shown in an image (displayed above). The depiction featured two children attempting to steal cookies from behind their mother’s back as she washes the dishes, with the sink overflowing. The participants wrote their responses, and the researchers utilized machine-learning artificial intelligence (AI) to analyze the verbal complexity of their submissions. The AI program quickly picked up on subtle differences criteria such as word choice, repetitions, and verbosity, and the scientists used statistical classification to determine the potential onset of AD .
The results were remarkable, especially considering the study’s minimized approach. The AI correctly predicted the onset of AD with 74% accuracy across all participants. Among certain demographics, the program performed even better, predicting the progression of AD in female participants with 83% accuracy compared to just 64% for males .
While more testing needs to be conducted, this study introduces a breakthrough in using language to detect the development of AD. Incorporating such a straightforward approach, the method satisfies the need for cheap, yet reliable, indicators of the disease, which are often much simpler and can be easily made widely available.
Another relevant example of innovation in AD diagnosis involves blood tests. A July 2020 study published in The Journal of the American Medical Association developed a test that measures levels of the protein p-tau217 in the brain and correctly diagnosed the existence of AD within participants an incredible 96% of the time . However, what sets the IBM investigation apart from other straightforward techniques is its predictive ability. Unlike other studies, all of the IBM participants started as cognitively normal, experiencing no, not even mild, symptoms of mental decline .
“This is the first report I have seen that took people who are completely normal and predicted with some accuracy who would have problems years later,” remarked Dr. Michael Weiner at the University of California, San Francisco . The average time for those who developed AD between participation in the study to eventual diagnosis was 7.59 years .
However, if there is no cure for AD, why would it be important to predict it? The FDA has approved drugs that can alleviate specific AD symptoms such as memory loss . Doctors can also recommend lifestyle changes such as physical activity and reducing alcohol/tobacco usage, all of which are associated with improving cognitive function . While these cannot deter or inhibit the disease’s progression, they can help patients and their providers by improving quality of life through early detection.
As breakthroughs emerge, predictive tests will become essential. An experimental drug manufactured by the pharmaceutical company Eli Lilly completed Phase 2 trials last month, demonstrating a 32% deceleration in cognitive decline in participants who received the drug . Simply put, identifying a disease as early on as possible will open up possibilities for more timely and effective treatment in the future.
While more experiments should be conducted, this simple language test presents an easily-accessible possibility for predicting the development of a debilitating disease. This approach fuses established clinical methods with AI technology to incorporate the best of both worlds. Before long, those who suffer from AD may be able to take their health and livelihoods back into their hands and confront this crippling condition head-on as a result of the vital research towards early diagnosis happening today.
 The Alzheimer’s Association. 2020 Alzheimer’s Disease Facts and Figures [Internet] [Cited 2021 Feb 14]. Available from: https://www.alz.org/alzheimers-dementia/facts-figures
 Straiton J. Predicting Alzheimer’s disease. BioTechniques [Internet]. 2019 [Cited 2021 Feb 15]; 67(4), 146-148. DOI: 10.2144/btn-2019-0114
 Eyigoz E, Mathur S, Santamaria M, Cecchi G, Naylor M. Linguistic markers predict onset of Alzheimer's disease. EClinical Medicine [Internet]. 2020 [Cited 2021 Feb 14]; 28. DOI: 10.1016/j.eclinm.2020.100583
 Palmqvist S, Janelidze S, Quiroz YT, Zetterberg H, Lopera F, Stomrud E, et. al. Discriminative Accuracy of Plasma Phospho-tau217 for Alzheimer Disease vs Other Neurodegenerative Disorders. JAMA [Internet]. 2020 [Cited 2021 Feb 21]; 324(8), 772-781. DOI: 10.1001/jama.2020.12134
 Kolata G. Alzheimer’s Prediction May Be Found in Writing Tests. The New York Times [Internet]. 2021 Feb 1 [Cited 2021 Feb 20]. Available from: https://www.nytimes.com/2021/02/01/health/alzheimers-prediction-speech.html
 The Mayo Clinic. Alzheimer's: Drugs help manage symptoms [Internet] [Cited 2021 Feb 14]. Available from: https://www.mayoclinic.org/diseases-conditions/alzheimers-disease/in-depth/alzheimers/art-20048103
 National Institutes of Health. Combination of healthy lifestyle traits may substantially reduce Alzheimer’s [Internet] [Cited 2021 Feb 15]. Available from: https://www.nih.gov/news-events/news-releases/combination-healthy-lifestyle-traits-may-substantially-reduce-alzheimers
 Kolata G. Alzheimer’s Drug Shows Promise in Small Trial. The New York Times [Internet]. 2021 Jan 11 [Cited 2021 Feb 20]. Available from: https://www.nytimes.com/2021/01/11/health/alzheimers-amyloid-lilly.html?action=click&module=RelatedLinks&pgtype=Article
[Image] Fors KL. The Cookie-theft picture [Internet] [Cited 2021 Feb 28]. Available from: https://www.researchgate.net/figure/The-Cookie-theft-picture_fig1_317095410
Leave a Reply.