Health and social care

AI is transforming all aspects of healthcare diagnoses, treatment, and health and care delivery to be more efficient and cost-effective.

AI and pattern recognition are being used to detect diseases such as cancer earlier and helping clinicians to take a more comprehensive approach to diseases and coordinate better care-plans for patients.

Find out how UCL researchers are using AI to solve health and social care issues below.

Big data

BigData@Heart

BigData@Heart will integrate healthcare data, activity monitors (wearables), state-of-the-art ‘-omics’ profiles, information about patients' lifestyles and health and their own reporting of symptoms, to better understand the causes of these conditions and the different subtypes, using innovative statistical, machine learning and data-mining methodologies.

CALIBER

CALIBER is a unique research platform consisting of ‘research ready’ variables extracted from linked EHRs from primary care, coded hospital records, social deprivation information and cause-specific mortality data in England.

Digital behaviour change interventions

Understanding incentives for increasing physical activity in similar groups of people and using machine learning to identify patterns in tracked activity data, to optimise outcomes and improve the matching of rewards and behaviour over time.

Exploring large datasets

Machine learning methods exploring large datasets, particularly from the National Hospital for Neurology and Neurosurgery (NHNN), to understand anatomical variability and pathological presentation, augmented by diagnostic and radiological report data.

Health Data Research UK

Health Data Research UK (HDR-UK) will include a London Research Data Warehouse with a specialised GPU cluster for specialist applications including AI. HDR-UK researchers are applying AI for medical treatment and diagnosis using Electronic Health Records (EHRs) to support medical decisions by cross-referencing complex data, finding correlations and considering seemingly unrelated external factors. One of three UKRI Innovation/Rutherford Fellowships are awarded to the Institute of Health Informatics (IHI), as part of the HDR-UK Institute.

Cancer

Deciphering intra-tumoural T cell receptor diversity

Developing mathematical and computational tools to decipher intra-tumoural T cell receptor diversity in non-small cell lung cancer and its relation to intra-tumour heterogeneity and clinical outcomes.

Developing advanced radiotherapy technologies

Development of novel imaging techniques, instrumentation and computer simulation. The Advanced Radiotherapy Technologies Network (ART-NET) is a Cancer Research UK (CRUK) Accelerator that brings together experts from across the UK to develop, assess and implement advanced radiotherapy technologies, utilising data from radiotherapy patients, AI/machine learning/deep learning computation and modelling, and UCL's data repository.

Enhancing the discovery of solid cancers

Quantitative methods of tissue and cellular imaging analysis in molecular pathology to enhance the discovery of diagnostic and therapeutic biomarkers in solid cancers.

Identifying personalised treatments for cancer patients

Cancer is a particularly challenging disease as tumours evolve in response to treatment. Successful outcomes often require sequences of treatment combinations over time. Cancer is also highly personalised. How a tumour develops depends also on the patient's genetics and the environment. For these reasons, cancer brings particular AI challenges. It requires developments in machine learning (ML) to deal with the heterogeneity of data across patients' treatments and outcomes. To progress and develop new therapies it is necessary to gain a better understanding of the mechanisms underlying the disease. This requires a focus on ML techniques that aid interpretability. Harnessing the power of AI and ML to study cancer evolution and the mechanisms of drug resistance to develop new diagnostic techniques and kinder, more effective therapies for cancer patients.

Revealing non-invasive markers of cancer pathology

Computational medical imaging of tissue microstructure and using VERDICT to develop MRI methods to reveal non-invasive markers of cancer pathology.

Studying genomic instability and immune activity in cancer

Computational methods to study the evolution of genomic instability and immune activity during cancer development and progression.

Drug discovery

Using computational methodologies to aid drug discovery

Applying novel computational methodologies to investigate complex biological problems including unsupervised machine learning adaptive sampling methods in molecular dynamics simulations, handling big data, dimensionality reduction methods among others, to aid drug discovery.

Healthcare modelling

High-dimensional modelling of imaging and clinical data

High-dimensional predictive, inferential, and prescriptive modelling of routine imaging and clinical data in acute neurology, for clinical and operational tasks.

Integrative approaches to complex biological and medical phenomena

Integrated Connectedness for a New Representation of Biology (ICON-BIO) aims to design holistic, integrative approaches to modelling and analysing the complexity of biological and medical phenomena, with a focus on redefining currently accepted paradigms in data science, biology and medicine, to find new treatments for complex and currently incurable diseases.

Infectious diseases

Tracking flu in real-time

Using symptoms of influenza-like illness reported on social media and via web searches to develop models for estimating flu to help track potential outbreaks.

Using AI to tackle the SARS-CoV-2 virus and associated COVID-19 disease

Developing machine learning, deep learning and AI techniques to accelerate the development of antiviral drugs fro COVID-19 by modelling proteins that play critical roles in the virus life cycle, in order to identify promising drug targets.

Neuroimaging

Detecting abnormality in epilepsy

Detection of the underlying abnormality in epilepsy patients with normal conventional MRI scans using novel contrasts and machine learning techniques.

Developing new MRI techniques

Mathematical and computational modelling with machine learning and parameter estimation techniques to engineer new MRI techniques for conditions such as Alzheimer’s disease, multiple sclerosis, and psychiatric disorders.

Improving dementia clinical trials

Improving dementia clinical trials and diagnostics using MRI scanning.

New methods for diagnosis

Developing new methods for diagnosis, such as using machine learning methods for automated analysis of the EEG, or using EIT to assist in a project to treat many conditions by electrical stimulation of autonomic nerves.

Pattern Recognition for Neuroimaging Toolbox

The Pattern Recognition for Neuroimaging Toolbox (PRoNTo) is based on machine learning techniques for the analysis of neuroimaging data. PRoNTo aims to facilitate the interaction between machine learning and neuroimaging communities.

Ophthalmology

Using AI to detect eye disease

Exploring how machine learning can help medical research into eye diseases. Researchers at Moorfields Eye Hospital NHS Foundation Trust, Institute of Ophthalmology and DeepMind Health have developed an AI system that can recommend the correct referral decision as accurately as experts for more than 50 eye diseases.

Surgery

Computer Assisted Navigation, Diagnosis and Intervention

The Computer Assisted Navigation, Diagnosis and Intervention (CANDI) research group applies artificial intelligence, computational modelling and software engineering on surgical procedures and medical interventions. CANDI aims to translate its methodologies and innovations to improve surgical navigation, disease diagnosis and minimally invasive interventions.