Researchers screened the genomes of thousands of individuals in an effort to identify genes linked to Alzheimer’s disease. But these scientists encountered a major obstacle: it is difficult to know with certainty which of these people have Alzheimer’s disease. There is no foolproof blood test to detect the disease, and dementia, a key symptom of Alzheimer’s, is also caused by other disorders. Early-stage Alzheimer’s disease may not cause any symptoms.
Now, researchers have developed artificial intelligence (AI)-based approaches that could be useful. An algorithm efficiently sorts through a large number of brain images and selects those that exhibit characteristics of Alzheimer’s disease. A second machine learning method identifies important structural features of the brain — an effort that could eventually help scientists detect new signs of Alzheimer’s disease in brain scans.
Overcoming Alzheimer’s disease: a look at the therapies of the future
The goal is to use people’s brain images as visual “biomarkers” of Alzheimer’s disease. Applying the method to large databases that also include medical information and genetic data, such as the UK Biobank, could allow scientists to identify genes that contribute to disease. In turn, this work could help create treatments and models to predict who is at risk of developing the disease.
Combining genomics, brain imaging and AI allows researchers to “find brain measurements that are closely linked to a genomic driver,” says Paul Thompson, a neuroscientist at the University of Southern California in Los Angeles. who is leading the efforts to develop these algorithms.
Thompson and others described the new AI techniques Nov. 4 at the American Society of Human Genetics annual conference in Washington DC.
Overwhelmed with data
Thousands of people have had their genomes sequenced and their brains analyzed over the past two decades as part of efforts to create massive research databases. But the speed at which this torrent of information is produced exceeds the ability of researchers to analyze and interpret it.
“We’re very data rich these days compared to how things were 5-10 years ago, and that’s where AI (and machine learning) approaches can excel,” says Alison Goate, geneticist at the Icahn School of Medicine at Mount Sinai. At New York.
In 2020, Thompson launched AI4AD, a consortium of researchers across the United States that aims to develop AI tools to analyze and integrate genetic, imaging, and cognitive data related to Alzheimer’s disease. As part of this project, researchers created an AI model trained on tens of thousands of magnetic resonance imaging (MRI) brain scans. These images had already been examined by doctors, who had selected scans showing signs of Alzheimer’s disease. From the images, the AI tool learned what the brains of people with and without Alzheimer’s disease look like.
In an essay, reported in a preprint1 which has not yet been peer-reviewed, the AI classifier detected Alzheimer’s disease in brain scans with over 90% accuracy. The consortium also used a similar approach to create a classifier that could accurately sort scans into distinct categories based on specific pathological changes in the brain associated with cognitive decline and dementia.2.
Degui Zhi, a data scientist at the University of Texas Health Science Center at Houston, and his colleagues took a different approach. While Thompson and his team focused the AI model on areas of the brain known to be linked to Alzheimer’s disease, Zhi wanted the tool to learn on its own structural features of the brain that can help diagnose disease.
How AI could lead to a better understanding of the brain
The researchers’ AI tool looked at thousands of brain scans and chose the features that most reliably differentiated one person’s brain from another.3. Zhi says this minimizes the likelihood of human bias influencing the algorithm. Today, Zhi’s team is using the algorithm to identify the traits that best distinguish brain scans of people with and without Alzheimer’s disease.
Thompson and Zhi acknowledge that AI models are only as good as the data they are trained on. There is a lack of racial and geographic diversity among individuals whose brains have been scanned and whose genomes have been sequenced, particularly in databases such as the UK Biobank, so the results of this research guided by AI might not be applicable to everyone. Additionally, Goate says it will be crucial to show that the performance of AI models can be replicated in other databases and that they show consistent results.
Rudolph Tanzi, a neurogeneticist at Massachusetts General Hospital in Boston, says these biomarkers could one day be part of a set of risk scores for the disease that also incorporate blood biomarkers and genetics. When all of this data is combined, risk scores can become “exponentially more sensitive,” which will hopefully allow people to seek early treatment before the disease progresses, he adds.
Alzheimer’s disease is just the beginning, Thompson says. If this approach works, it could also be applied to other diseases that have a physical presentation in brain imaging, he says.