There is no cure for Alzheimer’s disease. There are however preventative measures. New promising drugs have emerged in recent years that can help curb the condition’s progression. Although, the thing with these medicines is that the treatments have to be administered early in the course of the disease in order to do any good.
This battle against time is what inspired scientists to search for ways to diagnose the condition earlier. Jae Ho Sohn, MD, MS, a resident in the Department of Radiology and Biomedical Imaging at UC San Francisco says:
“One of the difficulties with Alzheimer’s disease is that by the time all the clinical symptoms manifest and we can make a definitive diagnosis, too many neurons have died, making it essentially irreversible.”
There’s good news. A test conducted by artificial intelligence has been created that could potentially give doctors a chance to intervene with treatment early. Researchers have programmed a machine-learning algorithm to diagnose early-stage Alzheimer’s disease, about six years before a clinical diagnosis is made, by using a common type of brain scan. Sohn combined neuroimaging with machine learning to try to predict whether or not a patient would develop Alzheimer’s disease when they first presented with a memory impairment – the best time to intervene. The study has been published recently in Radiology.
Positron emission tomography (PET) scans have been investigated as one tool to help diagnose Alzheimer’s disease before the symptoms become severe. PET scans measure the levels of specific molecules, like glucose, in the brain. There are also other types of PET scans that look for proteins specifically related to Alzheimer’s disease. The thing is, glucose PET scans are much more common and cheaper, especially in smaller health care facilities and developing countries since they’re also used for cancer staging.
Glucose is the primary source of fuel for brain cells. The more active a cell is, the more glucose it uses. On the contrary, as brain cells become diseased and die, they use less and, eventually, no glucose. Knowing this, radiologists use PET scans that measure glucose levels to try and detect Alzheimer’s Disease by looking for reduced glucose levels across the brain, especially in the frontal and parietal lobes of the brain. It seems like the perfect strategy… If it wasn’t for the fact that the disease is a very slow progressive disorder – meaning the changes in glucose are very subtle and so difficult to spot with the naked eye.
To solve this flaw in the strategy, Sohn applied a machine learning algorithm to PET scans to help diagnose early-stage Alzheimer’s disease more reliably. He did this by feeding the algorithm images to train it. The images came from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) – a massive public dataset of PET scans from patients who were eventually diagnosed with either Alzheimer’s disease, mild cognitive impairment or no disorder.
“This is an ideal application of deep learning because it is particularly strong at finding very subtle but diffuse processes. Human radiologists are really strong at identifying tiny focal finding like a brain tumor, but we struggle at detecting more slow, global changes. Given the strength of deep learning in this type of application, especially compared to humans, it seemed like a natural application.”
After a while of feeding the algorithm information, it began to learn on its own which features are important for predicting the diagnosis of Alzheimer’s disease and which are not. After training the A.I. with almost 2,000 scans, the scientists tested it on two novel datasets (scans the algorithm hadn’t been fed yet) to evaluate its performance.
It did a phenomenally good job correctly identifying 92 percent of the patients who developed Alzheimer’s disease in the first test set and 98 percent in the second test set. Not only that, but it made these correct predictions at a little more than six years before the patient received their final diagnosis.
To make sure this will work accurately on real people Sohn is going to test and calibrate the algorithm on larger, more diverse datasets from different hospitals and countries. He says:
“I believe this algorithm has the strong potential to be clinically relevant. However, before we can do that, we need to validate and calibrate the algorithm in a larger and more diverse patient cohort, ideally from different continents and various different types of settings.”
If the algorithm can withstand these tests as well as the first ones, Sohn believes it could be employed when a neurologist sees a patient at a memory clinic as a predictive and diagnostic tool for Alzheimer’s disease. It will be a great help to get the patient the treatments they need sooner.