Asaf Raza

Research Abstract

To build a generalized and reproducible deep learning model a large dataset needs to tune millions of model parameters. Because of the sensitivity of medical images and tight regulation, collecting huge datasets for deep learning model training is the key problem due to the sensitivity of medical imaging and strict regulations. Federated learning paradigm offers machine/deep learning models training on multisite datasets without exchanging the data which preserves the privacy of data. Our goal is to propose a collaborative federated learning framework that enables numerous medical institutions to use deep learning to screen for various diseases from various images without transferring patient data.