About Ellen Xu
Ellen developed a computer model using a subset of machine learning called a convolutional neural network to improve diagnosis of Kawasaki disease, a leading cause of acquired heart disease and coronary artery aneurysms in children between the ages of one and five.
A Deep Convolutional Neural Network for Kawasaki Disease Diagnosis
View Poster
Ellen Xu, 17, of San Diego, developed a convolutional neural network, a subset of machine learning used for image analysis, to help diagnose the disease for her Regeneron Science Talent Search medicine and health project. As the leading cause of acquired heart disease and coronary artery aneurysms in children between the ages of 1 and 5, a KD diagnosis is based on five visual signs that can easily be confused with other diseases. To improve identification, Ellen pre-trained her model with images of KD and lookalike diseases from the Internet and images provided by parents of KD children. To generate a larger dataset, she altered the images by adding a variety of random photographic transformations. Ellen’s work indicates that her model can distinguish between children with and without clinical manifestations of KD with 85% specificity using a smartphone photo of the child.
The daughter of Lin Jiang and Hao Xu, Ellen is co-editor-in-chief of the Del Norte High School’s award-winning literary magazine, founder of a non-profit which has taught cybersecurity to girls of military families, and a nationally and internationally ranked saber fencer.
Beyond the Project
Ellen is first author of an article on her research published in the peer-reviewed journal Nature Scientific Reports.
FUN FACTS: Ellen designed and built a low-cost, fully functional, nine-foot-wide geodesic dome in her backyard over Thanksgiving break. To prepare, she studied geodesic dome calculations and stereometry and even printed a miniature 3D model.
