The challenge
The health care system's current approach to treating mental health often begins with symptoms. That’s valuable, but it’s only part of the picture.
Mental health and neurodevelopmental conditions can differ greatly from person to person.
Finding the right support can take time and involve trial and error. This process can be frustrating and distressing for both families and clinicians. This also means that chances for early intervention are often missed.
That’s where artificial intelligence can help.
Our goal is for a child to be identified early as being at risk for a mental health challenge - not just because of symptoms, but because of patterns in their data. They don't wait months. They walk into a clinic where a team uses objective tests to quickly understand what's going on. And from day one, their treatment is tailored to them. There's a feedback loop. If something isn't working, we adjust. If something is helping, we build on it. That's not a fantasy. It's where we're heading, and we're building it now.
Professor Gustavo Sudre, King's Maudsley Partnership
Our approach
At the King's Maudsley Partnership, we use AI to go beyond general diagnostic categories. This helps us gain a more precise and personalised understanding of each child's mental health.
Our researchers use large datasets that include brain imaging, genetics, and clinical records. They find patterns that aren't always clear by observing someone. Then, they use these patterns to develop better tools for clinicians, who in turn can make care more focused and timely.
AI tools depend on the data used to train them. Often, this data fails to represent the children who need help most. This includes children from underrepresented communities and those with severe symptoms, making it hard for them to complete research procedures.
Tools based on incomplete data may help some groups but can leave others out. This risks widening existing inequalities instead of reducing them.
This is why equity is a design principle in our work, not an afterthought. At the Pears Maudsley Centre, we’re investing in child-friendly brain imaging technology. This technology works well, even if children are anxious or can’t stay still.
We are building partnerships with schools and communities. This helps ensure that our research participants reflect the full diversity of the population we serve. We design our models with inclusion from the start. This way, the benefits of AI in children's research can help all children, not just those easy to study.
What we’re working on
Predicting ADHD outcomes using machine learning
In one of our most significant recent studies, we combined genetic and brain imaging data from children diagnosed with attention deficit hyperactivity disorder ADHD) and trained a machine learning model (a type of AI that learns patterns from large amounts of data) to predict their future outcomes.
The model can now predict with over 80% accuracy whether a child will continue to meet diagnostic criteria in adolescence, or whether symptoms will reduce over time. This level of insight can be transformative for families and clinicians trying to make informed decisions about support and intervention at the right moment.