Artificial Intelligence Predicts Dementia Several Years Ago with Just one Brain Scan

Artificial intelligence predicts dementia several years ago with just one brain scan and improves the quality of life of patients.

The University of Cambridge‘s AI will learn brain scans from thousands of Alzheimer’s patients and identify individuals with Alzheimer‘s disease in those patients who exhibit mild cognitive impairment, such as signs of memory loss or signs of language or visual/spatial perception problems. It was more than 80% accurate in making predictions, and was able to predict how quickly cognitive abilities would decline over time.

A groundbreaking study by Professor Zoe Kourtzi of the University of Cambridge, UK, has found that artificial intelligence (AI) can detect early signs of dementia years ago with a simple brain scan long before symptoms appear.

Dementia is characterized by the accumulation of various types of proteins in the brain, which damages brain tissue and causes cognitive decline. In Alzheimer‘s disease, these proteins include beta-amyloid, which aggregates between neurons to form ‘plaques’ that affect function, and tau, which accumulates inside neurons.

Molecular and cellular changes in the brain usually begin years before symptoms appear. Diagnosing dementia can take months or years. It usually requires two or three hospital visits and can be diagnosed using a variety of CT, PET, and MRI scans.

A team led by Professor Joy Kurchi, a psychology professor at the University of Cambridge and leading the Alan Turing Institute, has developed an artificial intelligence (AI) machine learning tool that can detect dementia in patients at an early stage.

The AI ​​has learned to use brain scans of thousands of Alzheimer’s disease patients to discover structural changes in the brain. The algorithm could also be combined with standard memory test results to provide a prognostic score, i.e. the likelihood that an individual will develop Alzheimer’s.

Notably, the AI ​​was more than 80% accurate in predicting individuals with Alzheimer’s disease in patients presenting with mild cognitive impairment, such as signs of memory loss or signs of language or visual/spatial perception problems, and over time They could even predict how quickly cognitive abilities would decline.

“We trained a machine learning algorithm to look for early signs of dementia by looking for patterns of gray matter loss (basically wear) in the brain,” said Kurchi. “We have been able to identify some patients with Alzheimer’s who have not yet developed symptoms, and this study will allow us to identify patients 5 to 10 years before symptoms appear as part of a health check-up over time.” I hope there is,” he said.

Although this AI algorithm has been optimized to look for signs of Alzheimer‘s, Kurchi and his colleagues are now learning to recognize different forms of dementia, each with its own unique pattern of volume loss in common.

Dr Timothy Rittman from the Department of Clinical Neuroscience and Addenbrooke’s Hospital at the Cambridge University Hospitals NHS Foundation are currently leading clinical trials to see if this approach is useful in a clinical setting. there is.

“We have shown that this approach works in a research setting,” said Dr Littman. “Now we have to test it in a real environment,” said Dr. Littman. “Two NHS trusts in the U.S. have been involved in clinical trials,” he said.

Dr. Littman said early detection of dementia is important for a patient’s quality of life and for a number of reasons. “It can be a very difficult period when a patient starts experiencing memory and cognitive problems. Being able to do it can give them clarity and put them at ease depending on their diagnosis or help them and their loved ones prepare for the long run.”

He continued, “If we detect the disease early enough, there are lifestyle changes we can recommend. For example, blood pressure medications, improved diet and exercise, and quitting smoking can help slow the progression of the disease. This could be,” he said.

Finally, the research team currently has very few drugs that can be used to treat dementia. One of the reasons clinical trials often fail is that once a patient develops symptoms, it can be too late to make any major changes.

Therefore, having the ability to identify individual lesions at a very early stage could help researchers develop new drugs. If this clinical trial is successful, the AI ​​algorithm can be applied to thousands more patients across the country, he said.

Meanwhile, the Cambridge University AI study that detects early signs of dementia through simple brain scans long before symptoms of dementia appear, was published based on the following two research papers.

 

In ‘NeuroImage (Jan. 26, 2020)’, an international renowned academic journal in the field of brain imaging, ‘Modeling prognostic trajectories of cognitive decline due to Alzheimer’s disease-down’ and peer evaluation As a non-predictive paper, it was published in the ‘Bioarchive (biorxiv.org/2020.8.17)’ with the title ‘Predicting future regional tau accumulation in asymptomatic and early Alzheimer’s disease-down’. each was published.

Source: https://www.biorxiv.org/content/10.1101/2020.08.15.252601v1.full.pdf