Decoding the brain: the developments of AI in Alzheimer’s disease research

When Alzheimer’s disease was first recognised in 1906, in a female patient displaying signs of confusion, aggression and paranoia, scientists were at a loss at how to treat this neurocognitive disorder; it was discovered during a period where patients with mental illnesses were still locked away in ‘lunatic asylums’, for there was no tangible infection or symptoms that could be treated with their current methods. It was worsened perhaps by the fact that symptoms of this disease were often inextricably linked to the natural effects of ageing – confusion, occasional memory loss and general weariness, and so patients with lesser symptoms of Alzheimer’s often went untreated at the very least and even unnoticed, and patients who experienced more obvious symptoms, such as aggression, mood swings and paranoia were institutionalised and removed from wider society. Alzheimer’s disease, despite medicine being a constantly evolving and improving field, had not experienced a breakthrough in research until very recently, with the increasing use of artificial intelligence (AI) in diagnostics and treatment. Whilst scientists have known for some time that Alzheimer’s is caused by the build-upRe of proteins amyloid and tau on the brain, diagnostics and treatment have been slow and challenging to develop. Furthermore, Alzheimer’s disease is best treated when discovered sooner, however symptoms remain inconspicuous in the early stages of the disease. However, AI has revolutionised this field. Previous diagnosis of Alzheimer’s relied almost solely on the physician's expertise and judgement, as patients’ cognitive abilities were scored via Mini-Mental State Examinations (MMSEs) and then diagnosis was confirmed via a brain scan, usually an MRI or a PET scan. However, these methods rely on the physician’s judgement, which is not standardised across hospitals, and due to the less distinct pathological features of Alzheimer’s in its early stages, brain imaging scans may lack the sensitivity required to accurately diagnose the condition. Alternatively, scientists are now using AI to analyse the development of Alzheimer’s across patients in order to create an accurate depiction of the timeline of the disease. For example, the Conditional Restricted Boltzmann Machine (CRBM) has analysed 18-month-long trajectories in 1,909 patients with mild cognitive impairment (MCI) or Alzheimer’s and was then able to simulate the disease trajectories of these patients, which could then be used to model the progression of Alzheimer’s disease. Research on early Alzheimer’s disease is mainly focused on MCI, due to it being the transition state between normal aging and Alzheimer’s. However, AI goes further than other methods of diagnostics and monitoring – it can analyse the time-series of a patient’s existing condition and then predict the changes in the patient’s cognitive function, rather than focusing only on the current state of the disease. Other teams of scientists are utilising AI as well. At the University of Cambridge, Professor Kourzi’s team have developed a machine-learning algorithm that has been programmed to spot structural changes in the brain. This, coupled with the MMSEs, was able to provide a prognostic score of the likelihood of an individual having Alzheimer’s disease. It was also proved to be over 80% accurate in predicting those individuals who went on to develop Alzheimer’s disease from presenting with MCI. Similar studies were carried out in UC San Francisco, where research teams found a way to predict Alzheimer’s disease up to 7 years before symptoms start to present, by analysing patient records using AI. Conditions that influenced the prediction were high cholesterol and, in women, osteoporosis. These separate studies are individual pieces in the puzzle of Alzheimer’s disease, and AI is clearly the way forward. Despite limitations to this method, like the risk of excluding patients that don’t fit inside the algorithm’s parameters, it is clear that it has the potential to be used in all areas of research and throughout the development of the disease, from early recognition, to diagnosis, to treatment. With the emerging sophistication of machine-learning algorithms, AI is the future of diagnostics in Alzheimer’s disease research and ultimately medicine as a whole.