A new imaging method aims to give doctors a precise, objective number for inflammation, replacing the current subjective scoring system that varies between radiologists. This matters because inflammation assessment is currently inconsistent. In conditions like axial spondyloarthritis (a form of inflammatory arthritis affecting the spine), radiologists look at MRI scans and make qualitative judgments about whether inflammation is present and how severe it is. These judgments differ between experts, making it hard to track disease progression or compare results across clinical trials. The researcher will develop two linked tools: first, a method to produce quantitative tissue property maps from standard MRI scans (no special equipment needed), and second, an AI algorithm trained by multiple radiologists to interpret those maps consistently. If successful, the approach could transform how inflammatory diseases are diagnosed and monitored across the NHS. Instead of subjective reports, clinicians would get consistent, repeatable measurements of inflammation severity and treatment response. The software will be made freely available for research initially, with commercialisation planned for wider clinical use. The method is designed to scale beyond spondyloarthritis to other inflammatory conditions.
View original technical description
Summary: Inaccurate/imprecise assessment of inflammation is a pervasive, fundamental problem in clinical care and trials. There is an unmet need for objective, quantitative imaging assessments of inflammation and structural damage. I will address this important problem by developing a new paradigm for inflammation assessment: Quantitative, Intelligent Inflammation Imaging (QIII), using axial spondyloarthritis (axSpA) as an exemplar. This new paradigm will use cutting-edge imaging methods to deliver consistent, quantitative assessments of inflammation severity, phenotype and course, using only widely-available acquisitions and without needing bespoke, specialised scans. The QIII approach will amalgamate experts' knowledge into generalisable computational imaging algorithms, thus 'democratising' inflammation imaging and delivering consistent, accurate inflammation assessment across the NHS. Plan Overview: The analysis plan has been developed in close collaboration with Sue Mallett, Professor in Diagnostic & Prognostic Medical Statistics (see also Detailed Plan). WP1: Overarching aim: To produce quantitative images showing inflammation without the confounding effects of non-inflammatory processes, I will develop a novel 'accessible quantitative MRI (qMRI)' method, producing state-of-the-art tissue property maps from widely-available, standard MRI acquisitions. Approach: The new method will build on the applicants' recent technical innovations; its performance will be assessed in MRI test objects and IDEAL participants' MRI scans, in terms of tissue property assessment accuracy, image quality and sensitivity/specificity for inflammation. WP2: Overarching aim: To make high-quality image interpretation more accessible, I will develop an AI algorithm that amalgamates multiple expert readers' expertise into a generalisable algorithm. Approach: A training subset of IDEAL MRI scans will be read and labelled by four radiologists; these labels will be used to train a convolutional neural network that learns both the spatial distribution of inflammation as well as binary assessments of the presence/absence of inflammation and individual features (oedema/fat/erosions) for each participant. The improvement in specificity/sensitivity for AI-assisted imaging interpretation compared to conventional image interpretation will be evaluated in a test subset of IDEAL scans for three further 'test' radiologists. WP3: Overarching aim: To enable quantitative, easily-interpretable measurement of treatment response and monitoring of disease severity, I will develop methods for precise measurement of temporal changes in inflammation. Approach: MRI scans acquired from patients before and after biologic therapy will be used to develop computational methods for assessing treatment response. The responsiveness (ability to measure treatment-related changes in inflammation) of the new methods will be compared against conventional MRI. WP4: Overarching aim: To facilitate the use of QIII in clinical workflows, I will develop and implement strategies that ensure the successful real-world application and integration of QIII in clinical settings. Approach: WP4 will employ a mixed-methods design, based on Normalisation Process Theory, incorporating workshops, surveys, and iterative feedback loops to engage stakeholders and facilitate real-world implementation. Impact: The QIII approach will transform the way that inflammation is imaged, starting with spondyloarthritis but with clear scalability across inflammatory diseases. QIII will enable objective, accurate diagnosis/monitoring/response evaluation in clinical care and provide robust endpoints for trials. The software will be made freely-available for research, and will ultimately be commercialised to enable widespread clinical and trial use.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know