Abstract
Intracranial hypertension (ICH) occurs when the pressure within the brain is elevated, typically due to a neurologic condition such as a brain tumor. Detection of ICH in children is complicated by the non-specific nature of symptoms, such as headaches. This leads to delays in diagnosis of ICH as well as unnecessary, costly, and sometimes invasive testing of children who are ultimately diagnosed with disorders other than ICH, such as migraines. There is a critical need for a non-invasive and cost-effective technique to screen children with symptoms of ICH and determine which patients require further investigations for neurologic conditions. Fundus photography is one promising solution. Newer technology has enabled fundus photographs to be captured using a portable, handheld device without eye drops.
Previous research, including a study by Drs. Chang and Narayanan (co-PIs), has shown that artificial intelligence (AI) applied to fundus photographs can accurately detect papilledema (optic nerve swelling due to ICH) in children. Because papilledema occurs in only ~20% of children with ICH, additional research is needed to develop a more generalizable AI algorithm.
In the proposed study, we will recruit 500 children (1,000 fundus photographs) to develop and test an AI model to detect ICH on fundus photographs even in the absence of papilledema (Aim 1). Additionally, we will use multiple regression to develop a formula to predict CSF pressure based on the AI model’s output (Aim 2). This single-site study will provide the foundation for a planned R01 multi-site study to determine the impact of AI interpretation of fundus photographs on rates of neuroimaging and time to diagnosis of ICH in children who present with symptoms of ICH. This research will be especially impactful in resource-limited settings such as community and rural hospitals, where access to an ophthalmologist and/or sedated MRI scans is limited.