Research interest: Deep Learning algorithms, Convolutional Neural Network, Generative Models and their applications in Computer Vision and Medical Image Analysis
Working with Mehran Ebrahimi.
My research involves developing deep learning techniques aimed at deformable image registration (alignment) problem applied to medical imaging applications. The long-term goal of my research is to incorporate these techniques into computer-aided diagnosis (CAD) systems to improve diagnostic accuracy and speed. Our contribution to this research is threefold: first, to design a novel convolutional neural networks (CNN) architecture that exploits the complex structure of medical images to predict spatial relationship between image pixels from a pair of images and directly estimate the displacement vector ﬁeld, second, a comparative study of conventional registration distance measures as the cost function, and third, effective data augmentation strategy to account for the scarce labelled data in the medical domain. My ultimate objective is to create an open source framework that leverages convolutional neural networks in medical image registration problem to empower high-performance CAD system development and enhance the quality of care for patients worldwide.