Skip to main content
Back to King's College London homepage

The Artificial Intelligence in Mental Health (AIM) lab is focused on understanding how to use artificial intelligence (AI) techniques to improve understanding of mental health disorders and improve diagnostic, prognostic, and treatment tools available for mental health clinicians.

The AIM lab is led Dr Paris Alexandros Lalousis (Lecturer in Artificial Intelligence in Mental Health) and Professor Nikolaos Koutsouleris (Chair of Precision Psychiatry). It is a research group based in the Department of Psychosis Studies at the Institute of Psychiatry, Psychology & Neuroscience at King’s College London

The underlying mechanisms of mental health disorders involve a complex mix of genetic, environmental, and biological factors. Individuals with the same diagnosis often have different underlying biological mechanisms and individuals with a different diagnosis often have similar biological mechanisms. This results in a difficulty to understand and predict illness presentations, response to treatment, and long-term functional outcomes.

In an effort to understand these mechanisms, several different types of data have been collected over the years including clinical, cognitive, environmental, blood, genetic and neuroimaging data. Recent technological and methodological advances, large international collaborations, and the availability Electronic Health Records and biobanks have further increased the volume of data available.

AI holds tremendous potential to transform clinical care by identifying patterns in data at the phenotypic, biological, and environmental levels. Our lab harnesses the power of various AI approaches for the most pressing questions in mental health care with an aim to translate data science into clinical care for transformative clinical applications.

People

Linda Bryant

PhD Student

Fiona Coutts

Research Associate in Transdiagnostic Machine Learning

Grace Jacobs

Postdoctoral Fellow

Nikolaos Koutsouleris

Professor (Professorial Chair) of Precision Psychiatry

Paris Alexandros  Lalousis

Lecturer in Artificial Intelligence in Mental Health

Maximin Lange

PhD Student

Projects

Health icons connected with lines against a light blue background
AIM Review

AIM Review is an web application with the aim of providing lightweight machine learning methods to speed up the labelling process in systematic reviews. AIM Review is a static web app, which means that it runs locally on your PC without connection to server and does not require any coding expertise or download any software, just your browser.

Diagram of a screen showing transdiagnostic biological clustering, represented by varying circles of different colours and sizes connected by lines
Transdiagnostic biological clustering

Individuals with different mental health diagnoses often share similar characteristics both in terms of their symptoms and their underlying biology as well as having a shared genetic underpinning. Conversely, individuals with the same psychiatric diagnoses often have differences in their symptoms, levels of inflammatory markers, and brain structure and function. This has detrimental effects on diagnosis, prognosis, and biomedical translation. The aim of this project is to identify biologically based phenotypes based on patterns of biological markers that may be present across different mental illnesses.

    White pill bottle spilling out varying medical drugs on a blue background
    Treatment response prediction across different disorders

    Treatment for psychiatric disorders takes on several forms including medication, psychosocial therapy, and, in extreme cases, electroconvulsive therapy. However there is no one treatment that is successful for all individuals: for example, about a third of people with psychosis do not respond to regular antipsychotics. It is not currently possible to predict whether an individual will respond to a medication or not, which leads to longer wait times for those who do not initially respond to find a successful treatment. Current research has found baseline brain, blood, and clinical biomarkers of treatment response. The aim of this project is to build prediction models using these biomarkers to predict treatment response, to enable the appropriate treatment to be selected in the first instance.

      Translucent diagram of the human brain, with an overlay of a network of dots and lines
      Neurobiologically relevant signatures for targeted drug development

      Current psychiatric medications come with several issues including lack of response in a percentage of the illness population, side effects, and not targeting all relevant symptoms. Therefore, alongside research in response prediction, there is also a search for novel drug targets that may alleviate more symptoms or have fewer side effects. The aim of this project is to find biological markers that are associated with specific aspects of psychopathology, and to investigate changes in these markers after treatment. These markers can then be used to develop novel medications that target different biological pathways compared to current medications.

        Two men in suits shaking hands
        King’s Algorithm for Acceptance Likelihood Identification (KAALI): Training a Machine Learning Model for Identifying Job Acceptance Likelihood for Individuals with Mental Health Conditions

        Individuals with mental illness (MI) experience unjust treatment in the labour market; contributing to, and facilitating symptoms of, their disease. Tackling this trend calls for a detailed understanding of mechanisms influencing hiring decisions and career trajectories, translating these findings into fruitful interventions for ensuring optimal, fair allocation of talent, and developments of methods for preventing wasted applications. Sufferers of MI of all severities are not only able to work and strive in competitive employment, but explicitly want to do so. (Mental) health and employment are heavily interrelated. Especially for individuals with MI, employment facilitates quality of life, illness management and recovery. Employment services have been established as an integral part of early intervention programmes (EIP) for MI, with the aim of employment regain and/or maintenance. Implementing such services in EIP ensures higher employment rates than programmes without. Playing a centre role in the age of digital recruitment is the use of Artificial Intelligence (AI) and Machine Learning (ML). This tool, when fully developed, will be able to limit time and increase the chances of successful job seeking for people with MI, and therefore actively play a part in early intervention, as well as management of their condition. This will not only allow millions of patients to alleviate suffering but save the UK government substantial amounts of money.

          People

          Linda Bryant

          PhD Student

          Fiona Coutts

          Research Associate in Transdiagnostic Machine Learning

          Grace Jacobs

          Postdoctoral Fellow

          Nikolaos Koutsouleris

          Professor (Professorial Chair) of Precision Psychiatry

          Paris Alexandros  Lalousis

          Lecturer in Artificial Intelligence in Mental Health

          Maximin Lange

          PhD Student

          Projects

          Health icons connected with lines against a light blue background
          AIM Review

          AIM Review is an web application with the aim of providing lightweight machine learning methods to speed up the labelling process in systematic reviews. AIM Review is a static web app, which means that it runs locally on your PC without connection to server and does not require any coding expertise or download any software, just your browser.

          Diagram of a screen showing transdiagnostic biological clustering, represented by varying circles of different colours and sizes connected by lines
          Transdiagnostic biological clustering

          Individuals with different mental health diagnoses often share similar characteristics both in terms of their symptoms and their underlying biology as well as having a shared genetic underpinning. Conversely, individuals with the same psychiatric diagnoses often have differences in their symptoms, levels of inflammatory markers, and brain structure and function. This has detrimental effects on diagnosis, prognosis, and biomedical translation. The aim of this project is to identify biologically based phenotypes based on patterns of biological markers that may be present across different mental illnesses.

            White pill bottle spilling out varying medical drugs on a blue background
            Treatment response prediction across different disorders

            Treatment for psychiatric disorders takes on several forms including medication, psychosocial therapy, and, in extreme cases, electroconvulsive therapy. However there is no one treatment that is successful for all individuals: for example, about a third of people with psychosis do not respond to regular antipsychotics. It is not currently possible to predict whether an individual will respond to a medication or not, which leads to longer wait times for those who do not initially respond to find a successful treatment. Current research has found baseline brain, blood, and clinical biomarkers of treatment response. The aim of this project is to build prediction models using these biomarkers to predict treatment response, to enable the appropriate treatment to be selected in the first instance.

              Translucent diagram of the human brain, with an overlay of a network of dots and lines
              Neurobiologically relevant signatures for targeted drug development

              Current psychiatric medications come with several issues including lack of response in a percentage of the illness population, side effects, and not targeting all relevant symptoms. Therefore, alongside research in response prediction, there is also a search for novel drug targets that may alleviate more symptoms or have fewer side effects. The aim of this project is to find biological markers that are associated with specific aspects of psychopathology, and to investigate changes in these markers after treatment. These markers can then be used to develop novel medications that target different biological pathways compared to current medications.

                Two men in suits shaking hands
                King’s Algorithm for Acceptance Likelihood Identification (KAALI): Training a Machine Learning Model for Identifying Job Acceptance Likelihood for Individuals with Mental Health Conditions

                Individuals with mental illness (MI) experience unjust treatment in the labour market; contributing to, and facilitating symptoms of, their disease. Tackling this trend calls for a detailed understanding of mechanisms influencing hiring decisions and career trajectories, translating these findings into fruitful interventions for ensuring optimal, fair allocation of talent, and developments of methods for preventing wasted applications. Sufferers of MI of all severities are not only able to work and strive in competitive employment, but explicitly want to do so. (Mental) health and employment are heavily interrelated. Especially for individuals with MI, employment facilitates quality of life, illness management and recovery. Employment services have been established as an integral part of early intervention programmes (EIP) for MI, with the aim of employment regain and/or maintenance. Implementing such services in EIP ensures higher employment rates than programmes without. Playing a centre role in the age of digital recruitment is the use of Artificial Intelligence (AI) and Machine Learning (ML). This tool, when fully developed, will be able to limit time and increase the chances of successful job seeking for people with MI, and therefore actively play a part in early intervention, as well as management of their condition. This will not only allow millions of patients to alleviate suffering but save the UK government substantial amounts of money.

                  AIM_logofinal

                  Group leads