Artificial intelligence (AI) heralds a new era in medical imaging as AI models recognize complex patterns in imaging data and provide quantitative assessment of imaging characteristics in disease diagnosis. Yet at present there are no AI tools to detect lesions caused by rheumatic diseases in clinical radiology. This is surprising given the high prevalence and heavy burden of these treatable diseases. Spondyloarthritis (SpA) affects 1% of the population but typically remains undiagnosed for 7 years after onset of symptoms, leading to irreversible joint damage in young adults. In this project I aim to translate AI from the computer lab into a clinical tool.
(Aim 1) Using AI I will develop a novel tool to detect unsuspected erosions and ankylosis of the sacroiliac joints in CT scans obtained from patients who are unaware that they have SpA. These hallmark lesions will be detected in an early stage of the disease by the tool and will then be confirmed by the radiologist.
(Aim 2) I will apply AI to discover a new MR imaging biomarker of early inflammation in SpA. The proposed new imaging biomarker will help diagnose SpA in the earliest stage of the disease, before changes become visible to the human eye, and will facilitate treatment monitoring and follow-up.
This project will pave the way for my long-term goal to translate AI to clinical practice in early detection of rheumatic disease and reach the highest impact on society: prevention of life-long disability in young patients.