Keywords: generative models, fundus images, latent space, patient attributes
Abstract: Screening for ophthalmic diseases routinely relies on retinal fundus images. These images are highly heterogeneous and little is known about how patient attributes such as age and ethnicity contribute to the variability in appearance. As the image variation due to such factors may ultimately confound automated image interpretation using deep learning models, understanding the influence of patient attributes on retinal fundus images is key for reliable AI applications in ophthalmology. Here, we draw on recent advances in generative modeling and present a population model of retinal fundus images which is capable of generating highly realistic images and allows for an analysis of how the patient attributes age and ethnicity are organized in the latent space of the generative model.
Registration: I acknowledge that acceptance of this work at MIDL requires at least one of the authors to register and present the work during the conference.
Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.
Paper Type: novel methodological ideas without extensive validation
Primary Subject Area: Application: Ophthalmology
Secondary Subject Area: Unsupervised Learning and Representation Learning
Confidentiality And Author Instructions: I read the call for papers and author instructions. I acknowledge that exceeding the page limit and/or altering the latex template can result in desk rejection.