Published on: May 2, 2018
by Women’s Brain Health Initiative:
Diagnosing Alzheimer’s disease (AD) is challenging, time consuming, and costly. Currently, there is no single test, or series of tests, that can determine with 100% certainty whether an individual has developed AD. In fact, AD cannot be definitively diagnosed until after death, when the brain can be closely examined for certain microscopic changes caused by the disease.
When an individual reports to a doctor that he or she has experienced bouts of memory loss or decreased cognitive function, he or she may be assessed using a variety of cognitive and physical tests, some quite invasive, to determine whether he or she “probably” has AD. However, this diagnosis requires visible symptoms that may only show up when it is too late to start preventative measures. This is of particular concern to researchers, many of whom are focusing their efforts on trying to find a treatment that will help prevent, slow, or reverse the disease. In order to develop such disease-modifying therapies, though, researchers require the participation of individuals who are at high risk of suffering from AD, or are in the very early stages. Finding individuals who fit this profile may now be easier thanks to artificial intelligence.
New artificial intelligence (AI) research suggests that doctors may soon have the tools to predict an individual’s likelihood of developing dementia several years before the onset of symptoms.
AI and Brain Scans
Canadian Researchers Combine AI and Amyloid PET Scans
Researchers from McGill University in Montreal, Canada used AI techniques to develop an algorithm that can, with a single amyloid positron-emission tomography (PET) scan, detect dementia signatures in the brains of individuals with mild cognitive impairment (MCI) two years before symptoms emerge. While scientists know that amyloid protein accumulates in the brains of individuals with MCI, not every patient who has MCI goes on to develop AD. Consequently, the presence of amyloid is not enough to determine if someone is in the early stages of AD.
This recent study — conducted by Mathotaarachchi et al. and published in Neurobiology of Aging in 2017 — made use of hundreds of amyloid PET scans obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), a global project in which participants complete a variety of imaging and clinical assessments to help scientists learn about the progression of AD. Using these amyloid PET scans, the researchers trained the AI software to identify which MCI patients would develop AD. There are subtle differences in the way that amyloid is distributed throughout the brain that the human eye cannot detect. However, the AI system was able to detect these differences, noticing patterns that distinguished between those who would go on to develop Alzheimer’s disease and those who would not. Remarkably, the algorithm was able to predict which MCI patients would later develop Alzheimer’s disease with 84% accuracy.
While the new software is currently available online for scientists and students, physicians will not be able to use it in clinical practice until it receives approval by health authorities. In the meantime, the researchers are conducting additional examinations to identify other biomarkers for dementia that could be incorporated into the algorithm to improve its prediction capabilities. Additionally, the McGill team is now testing the algorithm in different patient groups, including individuals who have existing conditions such as small strokes.
Researchers in Italy Train AI to Assess Brain Region Connectivity Using MRI Scans
A different team of researchers (Amoroso et al.) from the University of Bari in Italy has also trained an AI system using brain scans from the ADNI database. In this case, however, the researchers used 67 magnetic resonance imaging (MRI) scans (38 from individuals with AD and 29 from healthy individuals) to train the algorithm to correctly discriminate between diseased and healthy brains. After training the algorithm to correctly analyze the neuronal connectivity between different regions of the brain, the researchers tested the system on a new set of scans from 148 subjects. Of these, 52 were healthy individuals, 48 had AD, and 48 had MCI but were known to have developed AD between two and a half to nine years later.
The algorithm not only was able to distinguish between a healthy brain and one with AD with an accuracy of 86%, but also could differentiate between healthy brains and those with MCI 84% of the time. These findings were submitted to the Cornell University Library Medical Physics Archive in September 2017.
Researchers in the Netherlands Combine AI and Arterial Spin Labeling MRI Scans
A recent study conducted by scientists from VU University Medical Center in Amsterdam, published in the December 2016 issue of Radiology, applied AI learning methods to a special type of MRI called arterial spin labeling (ASL) to help diagnose AD in the early stages, particularly in centres that lack experienced neuroradiologists. This type of imaging is a non-invasive, quick, and increasingly widely-available method for quantifying blood flow in the brain. The researchers developed an algorithm that can distinguish between patients at various stages of AD with good to excellent accuracy. The study included 260 patients from the Alzheimer Centre at the VU University Medical Centre who underwent ASL MRI between October 2010 and November 2012. The system was able to distinguish effectively among participants with Alzheimer’s disease, MCI, and subjective cognitive decline (SCD). The researchers were then able to predict the Alzheimer’s diagnosis or progression of single patients with a high degree of accuracy, ranging from 82% to 90%.
AI and Speech
Artificial intelligence is also being used to assess and monitor Alzheimer’s disease and other types of dementia by simply examining a person’s voice. While memory impairment is the primary symptom of AD, language impairment usually occurs as well and therefore can be a good indicator of the severity of the disease over time. Recent research suggests that AI can be taught to accurately detect various cognitive disorders, including Alzheimer’s disease, using just short samples of speech.
Frank Rudzicz, a scientist at the Toronto Rehabilitation Institute and an assistant professor at the University of Toronto’s department of computer science, is one researcher studying how AI can be used with speech to detect AD. Rudzicz and his colleagues developed a test that takes only 45 seconds to analyze 400 different variables of speech and to predict the severity of AD with approximately 82% accuracy. Their AI models are currently being tested and trained to understand different languages and accents. For now, the tools are being used strictly to track cognitive decline in existing patients, not diagnose new patients.
Rudzicz pointed out that there are important ethical implications to consider before this type of technology could be used in medical practice. “Who will be held accountable for any misdiagnoses made by AI technology? How will patient privacy be protected? Will only health professionals have access to these tools or will they also be available to the general public? These are important questions that need to be answered before AI can be used ethically as part of the diagnosis or assessment of any disease, including Alzheimer’s.”
Source: MIND OVER MATTER V.6
While anyone can experience a stroke at any age, women experience more stroke events than men and are less likely to recover. “BE FAST” is a checklist of 6 items to keep in mind when assessing whether you might be...
Enjoy these highlights from our Engaging Millennial Minds Chew on This virtual event with celebrity chef Mark McEwan.
This virtual culinary event featured Celebrity Chef and Restauranteur Mark McEwan.
The material presented through the Think Tank feature on this website is in no way intended to replace professional medical care or attention by a qualified practitioner. WBHI strongly advises all questioners and viewers using this feature with health problems to consult a qualified physician, especially before starting any treatment. The materials provided on this website cannot and should not be used as a basis for diagnosis or choice of treatment. The materials are not exhaustive and cannot always respect all the most recent research in all areas of medicine.