A branch of artificial intelligence which is based on training computers to learn patterns has the ability to transform our understanding of dementia. David Llewellyn co-led the first DPUK datathon. He explains why machine learning could be the beginning of the end for this devastating condition.
Dementia isn’t a single disease and people with dementia, or who are at risk of dementia, vary enormously. Some forms of dementia are caused by neurodegeneration, including Alzheimer’s disease, and others are caused by vascular problems such as stroke. Not being able to diagnose it early enough is one of the biggest problems, and it’s in this area that AI techniques like machine learning give me hope.
Machine learning is a branch of artificial intelligence which relies on our ability to train computers to learn from data and identify hidden patterns. As computers have become more powerful, so has our ability to harness them and make sense of rich and large datasets. Machine learning already enhances our smartphones and makes internet searching efficient, and we suspect that machine learning has the potential to transform clinical medicine. Machine learning is particularly well suited to dealing with clinical data relating to complex conditions such as dementia, and for that reason we are excited to explore its potential in a datathon.
My hope is that machine learning will help us diagnose the disease early enough so people can receive better support and access ongoing research studies. Why haven’t we been able to diagnose it early? Simply put, the early stages of dementia are invisible. Clinicians are often unsure who to refer for costly, time-consuming and potentially worrying investigations at memory clinics. Some people never go to their doctor and around a third of cases are never diagnosed. As a result it is very challenging to assess patients and recruit them to trials that will be right for them. We need research that gives us new insights into the complex ‘disease signatures’ that underpin dementia and have the potential to enhance clinical practice. But where do we start? Many of my colleagues and I think machine learning is a very promising approach.
Not all machine learning scientists will be familiar with the datathon format but we’re already seeing enthusiasm within the community for this way of working. A datathon is an event where scientists come together to conduct analyses, challenge themselves, learn new techniques, and have fun. It’s a fantastic opportunity to meet other interesting people and make a difference to what is a really big societal challenge. My hope is that this will be the key to improving care, and fuelling the trials which we hope will result in new disease-modifying treatments.
What to read next
Although the more commonly known ‘hackathon’ has been used to address societal-level problems from homelessness to corruption, we think that it’s only recently that the data science community is turning its attention to dementia. The DPUK datathon series might well be one of the first for the field in fact. Chris, DPUK’s Data Project Manager, and Sarah, DPUK’s Senior Data Manager, take us through the three key ingredients for a successful datathon.