A0516
Title: ImgKnock: Knockoff inference with optic disc images for glaucoma diagnosis and risk factors
Authors: Zhe Fei - UC Riverside (United States) [presenting]
Abstract: ImgKnock is introduced, an innovative pipeline that leverages the power of knockoff inference and deep learning for the analysis of optic disc images in glaucoma. Knockoff inference, known for its ability to control false discovery rates in high-dimensional data, is adapted here to handle complex image data. ImgKnock uniquely generates knockoff images of fundus photographs, enabling the prediction of glaucoma-related outcomes and the identification and testing of important image regions for accurate diagnosis. The method extends knockoff generation to non-tabular data, maintaining key features like the swapping property and ensuring robust feature selection. Central to ImgKnock is a novel testing procedure based on deep neural networks (DNNs), which allows for controlled false discovery rates at the pixel level. This approach facilitates precise inference of feature importance, which is crucial for understanding the disease. ImgKnock has been successfully applied to various image datasets, including MNIST and CIFAR-10, demonstrating its versatility. Significantly, the analysis of optic disc images from glaucoma patients at UCLA Stein Eye Institute has led to novel clinical insights, pinpointing specific fundus regions that are predictive of glaucoma. This advancement in glaucoma research highlights ImgKnock's potential to contribute to improved diagnostic methods and a deeper understanding of the disease.
