Dr Sheng Chiong Hong from oDocs Eye Care has completed a study which shows it is possible for clinicians to develop artificial intelligence (AI) classifiers with no previous programming or AI knowledge, using standard laptop computers. A classifier is an algorithm that can be applied to a classification, or set of categories to which a new observation belongs.
Dr Hong and colleagues set out to develop an AI system that could detect several fundus pathologies and report relevant clinical features. To do this, they used 4,435 colour fundus photos obtained from publicly available fundus image databases. Images were uploaded to a web-based AI platform for training and validation of AI classifiers and they created separate classifiers for each fundus pathology and clinical feature.
In testing the system, the researchers looked at accuracy, sensitivity, specificity and area under a receiver operating characteristic curve (AUC) for each classifier. Acknowledging that statistical performance was limited by the small sample size, they reported average accuracy was 89%, average sensitivity was 75%, average specificity was 89% and average AUC was 0.58.
“This study is a proof-of-concept AI system that could be implemented within a diabetic photo-screening pathway. Performance was promising but not yet at the level that would be required for clinical application,” they wrote.
Dr Hong has demonstrated how clinicians can develop a deep learning algorithm without knowledge in coding and programming using MedicMind. The demonstration, How to build your own medical AI, is available on YouTube.
The study was published in Clinical & Experimental Ophthalmology May/June 2019.
