AI analyzes gray matter loss to predict the onset of Alzheimer's
Advances in artificial intelligence promise to open up all sorts of possibilities when it comes to health care, and analyzing medical images for signs of trouble is already proving to be a great strength of the technology. Scientists at Cambridge University have demonstrated a new type of machine learning algorithm they say can detect structural changes in the brain that are indicative of early dementia, and can be combined with standard memory tests to calculate the likelihood of someone going on to develop Alzheimer's.
While clinicians are well adept an analyzing PETs, MRIs or other types of medical images for irregularities that might be linked to disease, artificial intelligence promises to supercharge these forms of diagnostics. The power of modern computing makes machine learning algorithms very well suited to spotting subtle changes in say, brain tissue, that would escape even the highly trained eyes of today's physicians.
We've seen this technology show great promise in detecting ventricular dysfunction, a major precursor to heart failure, for example, or more efficiently detecting malignant tissue in the lung nodules that might be a sign of cancer. Along with these advances, we've also seen scientists make inroads in detecting the onset of Alzheimer's, potentially long before any symptoms appear.
A compelling example of this came in 2018, when University College London scientists demonstrated a new machine learning algorithm that could pick up subtle patterns in dense brain imaging data that represent changes in glucose uptake. Tested on a small set of brain scans, the algorithm was able to then predict every single case that would develop Alzheimer's an average of six years ahead of their diagnosis.
The Cambridge University scientists have been pursuing a similar possibility, but through a different physiological mechanism. The team trained their machine learning algorithm on brain scans of patients who went on to develop Alzheimer's, through which it learnt to detect structural changes related to gray matter density in the brain that denoted the formative stages of the disease.
“We have trained machine learning algorithms to spot very early signs of dementia just by looking for patterns of gray matter loss – essentially, wearing away – in the brain," says Professor Zoe Kourtzi. "When we combine this with standard memory tests, we can predict whether an individual will show slower or faster decline in their cognition. We’ve even been able to identify some patients who were not yet showing any symptoms, but went on to develop Alzheimer’s."
The algorithm was applied to brain scans of patients already with mild cognitive impairment and suffering from memory loss or trouble with language, visual or spatial perception as a result. Combined with the memory tests, the algorithm proved capable of predicting who would go on to develop Alzheimer's with 80 percent accuracy, and was also able to predict the rate of their cognitive decline.
"In time, we hope to be able to identify patients as early as five to 10 years before they show symptoms as part of a health check.” says Kourtzi.
Like the other examples mentioned above, it is very early days for this research and it will require a lot more work to get it into clinical use, beginning with replicating the results in a larger cohort than the 80 patients involved in this initial research project. The scientists are now drawing up a clinical study to that effect, and while the algorithm is optimized for Alzheimer's, they hope to also adapt it to detect other forms of dementia by identifying other unique patterns of structural changes in the brain.
“We’ve shown that this approach works in a research setting – we now need to test it in a ‘real world’ setting,” says study author Dr Timothy Rittman.
The team's research paper, which is not yet peer-reviewed, is available here.
Source: Cambridge University