Computed Tomography (CT) reconstructs multi-dimensional images from a finite number of low dimensional projections. Used extensively in medical imaging for diagnosis and pre/post-surgical assessment , these images require ionizing radiation, which adds non-trivial clinical risk. Exposure to this radiation is directly correlated with the amount of data collected. This project aims to answer this question: Can we create a method to convert 1-D measurements to 2-D images using sparsely collected data, i.e. data requiring minimal radiation? And if that is possible, can we further push our algorithms, using the foundations of our previous question, to create a 3-D model from 2-D datasets?

I am participating in SuperUROP to gain research exposure, more specifically in medical research. My background as a computer science major gives me some insight into how to tackle my research problem. I’m excited to potentially make a real world impact with my research, and learn more about the possibility of using deep learning for medical imaging!