Video Short: Bringing clarity to Uveitis Diagnosis with AI

Up, up and away – clearing the clouds… Bringing clarity to Uveitis Diagnosis with AI: It’s a pleasure to share this joint Nature EYE editorial with Peter Thomas in which we survey  how artificial neural networks (ANNs) might be used to enhance the diagnosis of uveitis, a significant cause of vision loss worldwide: https://www.nature.com/articles/s41433-024-03588-y The Challenge: Uveitis, an intraocular inflammatory condition, presents a complex diagnostic puzzle due to its myriad causes and varied clinical presentations. Traditional diagnostic methods, often reliant on subjective clinical judgment, can lead to delays and inaccuracies. Collaboration: The partnership of data scientists and clinicians holds the key to unlocking the full potential of AI and machine learning in ophthalmology.  We review the state of play to accompany this paper from Jacquot et al (https://www.nature.com/articles/s41433-024-03530-2) who have used a multilayer perceptron (MLP) model, to predict the underlying causes of uveitis. Their model, trained on a dataset of 375 patients, demonstrated impressive accuracy, matching expert opinion in 77.8% of cases and achieving 93% accuracy within the top two probable diagnoses. We’re striving towards a more objective and standardised approach to uveitis diagnosis, and these models can help – particularly with diagnostic decision support systems (DDSS) which can complement (but not replace) clinicians – hopefully streamlining the clinical workflow and reducing diagnostic errors. The DDSS in the paper we feature focuses on diagnostically challenging cases (ideally providing valuable support where it's most needed…) While challenges remain in terms of geographic transferability, under-representation of rare aetiologies, and integration into electronic medical records, the potential of this technology is undeniable. Future research will focus on validating these tools across diverse populations and refining them for seamless integration into clinical practice.