PROJECT SUMMARY:
This application seeks a career development award for an academic vitreoretinal surgeon with an interest in high
myopia, a condition which predisposes patients to potentially blinding complications including retinal tears (RTs)
and rhegmatogenous retinal detachments (RRDs). This proposal is a 5-year curriculum and research plan to
transition Dr. Cassie Ludwig to independence. The candidate is an accomplished early career physician-scientist
who will undergo all training and execute the research noted herein during this period.
Myopia affects one third of the world's population today and has been predicted to affect 50% of the world's
population by 2050.1,2 Worse, this prediction is likely an underestimate as myopigenic behaviors have been
further compounded by the COVID pandemic and digital remote learning.3–8 This increasing prevalence has
significant consequences as each diopter of myopia increases the risk of retinal tears and detachments, myopic
macular degeneration, choroidal neovascularization, myopic traction maculopathy, strabismus, glaucoma, and
cataracts. Slowing myopia progression even minimally can help prevent blindness. Using combined data from
five large population-based studies, Bullimore et al. found that slowing myopia by one diopter should reduce the
likelihood of a patient developing an RRD by 30%.2
Electronic health records (EHRs) and ophthalmic imaging databases contain enormous quantities of systemic
and ocular data generated by clinical practice which can be used to better understand the relationship between
systemic and ophthalmic risk factors, myopia and RTs and RRDs. EHR and imaging data can be fused into
predictive models that employ machine learning to risk-stratify patients.
In this proposal, Dr. Ludwig aims to achieve the following: 1. Develop and validate a structured EHR deep
learning framework to predict RT and RRD risk in myopes and non-myopes 2. Develop and validate an
unstructured EHR transformer-based deep learning model to predict RT and RRD risk in myopes and
non-myopes, and 3. Develop and validate an ultra-widefield photography convolutional neural network
(CNN)-based deep learning model to predict RT and RRD risk in myopes and non-myopes. The central
hypothesis is that modeling of attributes from EHR data and images can predict risk of RTs and RRDs.
The principal investigator, Cassie A. Ludwig, MD, MS, will perform this research as part of a larger effort to obtain
additional training and mentorship in biomedical informatics, artificial intelligence, biodesign, and myopia. Dr.
Ludwig’s career development plan includes a PhD program with didactic coursework, conferences, workshops,
and frequent communication and interaction with a network of mentors with an impressive abundance of their
own NIH funding and prior mentorship experiences. This experience will guide Dr. Ludwig into a career as an
independent clinician-scientist with expertise in artificial intelligence and a focus on myopia and its sequelae.