Artifical Intelligence (AI) has been programmed to use magnetic resonance imaging (MRI) scans to recognise the brains of people suffering from Alzheimer’s disease. Using these scans, it can even predict whether someone is likely to develop the condition a decade before the onset of symptoms.
Alzheimer’s disease is the most common cause of dementia, which affects 850,000 people across the UK and around 70,000 people in Scotland. This condition causes a gradual cognitive decline that manifests as confusion, memory loss and personality changes.
While it is known that amyloid and tau proteins play a role in the disease progression, our understanding of Alzheimer’s is incomplete and treatment options are limited, especially after symptoms appear.
Currently, we can only measure the amount of these proteins in the brain (which is related to disease severity) using invasive and expensive techniques.
This new research, led by Nicola Amoroso and Marianna La Rocca at the University of Bari, could be a big leap forward. It is hoped the AI will increase safety and decrease the cost of disease prediction.
The AI’s prediction algorithms are based computational patterns modelled after the human brain and nervous system. It is therefore fitting that these are being used to diagnose a neurological condition.
These prediction algorithms consist of two steps. In the first step, called the training step, the AI learned to tell the difference between healthy and diseased brain scans.
To make these distinctions, the AI looked for differences in structural connections in small regions of healthy and diseased brain scans.
In the second step, called the validation or testing step, the trained AI attempted to classify new scans without having access to their true labels. Two thirds of these scans were of brains that were either healthy or diseased, while the remaining third were of brains showing signs of mild cognitive impairment (MCI) that led to Alzheimer’s disease two and a half to nine years later.
The AI was able to tell the difference between healthy and Alzheimer’s brains 86 per cent of the time, and it also made the distinction between healthy and MCI brains 84 per cent of the time.
This has important implications for disease prevention and management, as early diagnosis and treatment could slow or stop the progress of Alzheimer’s.
The results of this AI research were published just in time for World Alzheimer’s Day on the 21 September, and also coincided with the announcement of a £600,000 award granted by Alzheimer’s Research UK (the largest charity funder of dementia research in the UK) to three Scottish research projects.
One of these projects is led by Dr Riccardo Marioni, a University Chancellor’s Fellow based at the Centre for Genomic and Experimental Medicine and the Centre for Cognitive Ageing and Cognitive Epidemiology.
Dr Marioni said: “In this project we will use novel statistical approaches to integrate genetic, epigenetic, transcriptomic, and health data to better understand the causes of Alzheimer’s disease and improve disease prediction.”
The results obtained from AI and scientists alike will improve our understanding of Alzheimer’s disease and ultimately expand our options for its treatment and prevention.