Oxford Clinical Artificial Intelligence Research (OxCAIR)

Oxford Clinical Artificial Intelligence Research (OxCAIR) is a multidisciplinary research group based in Oxford University Hospitals (OUH).

OxCAIR focuses primarily on evaluating digital health technologies that have received regulatory approval or are commercially available, with a particular interest in artificial intelligence (AI)-assisted image analysis applications.

The OxCAIR group, which comprises a dedicated team of senior academics, fellows and administrative staff, works on a broad portfolio of projects, often in partnership with clinical, academic and industrial stakeholders in Oxford, across the UK and internationally.

The group has secured more than £3 million funding for a diverse portfolio of AI evaluation projects, as well as being selected to host the NHS Clinical AI Fellowship programme in the Thames Valley region.

Aims of OxCAIR

Undertake detailed and robust evaluation of mature/commercial AI-led technologies which aim to assist or undertake clinical activity, assessing their performance, usefulness and reliability
Accelerate the adoption of AI in healthcare by creating a pipeline for generating high-quality clinical evidence
Bridge the gap between AI innovation and its implementation in the NHS, leveraging Oxford's expertise and the NHS's unique position to conduct large-scale studies
Develop an enhanced methodological approach for assessing clinical AI technologies
Foster collaboration between academia, industry and healthcare providers to enhance patient care, improve NHS efficiency and establish Oxford and the UK as a global leader in clinical AI research and implementation
Develop the skills and understanding of clinicians and academics at OUH and beyond with respect to the clinical evaluation of AI-led technologies

Projects

SAMURAI Fracture

Systematic Assessment of Medical Utility of AI-assisted Fracture Detection

This study will evaluate whether using AI software to help clinicians detect broken bones on X-rays can improve patient care in emergency departments and minor injury units.

Misdiagnosing fractures is a common issue that can lead to unnecessary visits, delayed healing and increased NHS costs. AI tools may reduce these errors, but it is unclear if they improve patient outcomes or save money.

This project has been awarded just under £400,000 by Small Business Research Initiative (SBRI) Healthcare to test AI software developed by Radiobotics. The OxCAIR team will collect data from several UK hospitals and minor injuries units where the technology is being trialled.

Accelerating Trustworthy AI

The OxCAIR team is part of a consortium comprising industry, NHS and academic organisations who have received £1.6 million from Innovate UK to evaluate the accuracy and effectiveness of a series of AI tools, including the following.

An obstetric measurement tool
Several AI algorithms for identifying pneumothorax, comparing them with a retrospective dataset of chest X-ray images
Several AI algorithms for identifying fractures, comparing their diagnostic accuracy with a retrospective dataset of images
The Jiva RDX AI tool for prostate cancer diagnosis within the Thames Valley and Surrey Secure Data Environment (SDE): the project involves reviewing and analysing prostate MRI scans and related patient data to assess the AI tool's performance
The Qure CXR algorithm for assessing chest X-rays. This evaluation will be done at OUH in 'shadow mode', meaning it takes place in the background and doesn't affect routine clinical work. These projects are being conducted using technology developed by the OUH spin-out company RAIQC

AI-REACT Multicase Multi-reader study and ACCEPT-AI trial

These two studies are evaluating the impact of Qure AI's qER EU 2.0AI tool, which is used alongside head CT scans.

AI-REACT is looking at how different healthcare professionals interpret the scans both with and without the assistance of the AI tool.

ACCEPT-AI is a multicentre trial which evaluates the ability of qER EU 2.0 to identify critical abnormalities on CT head scans and flag them for urgent radiology reporting.

OxCAIR has received a total of £487,000 from NHS Digital (formerly NHSX) to evaluate Qure AI's head CT algorithm tool in these two studies.

LUNIT CXR Muliticase Multi-reader Study

This study is evaluating the performance of a deep-learning AI-based model, Lunit INSIGHT CXR, in assisting clinical staff to interpret chest X-rays and make clinical management decisions.

FRACT AI Multicase Multi-reader Study

The study aims to evaluate the diagnostic accuracy of the Gleamer Boneview algorithm and its impact on the diagnostic performance of frontline clinicians in detecting fractures on plain radiographs.

Simulation Training in Emergency Department Imaging 2 (STEDI2)

This multicentre trial is assessing the impact of an online training simulation on Emergency Department (ED) clinicians' ability to interpret head CT images encountered in routine clinical practice.

This project has been supported with £800,000 in funding from the Small Business Research Initiative (SBRI).