BISPL Lab Introduction

by Jong Chul Ye.

The Bio Imaging, Signal Processing & Learning (BISPL) Lab is a research group in the department of Bio & Brain Engineering at KAIST dedicated to investigating the beauty of biomedical imaging with the help of mathematics, machine learning, and physics. We currently have 17 Ph.D. and Masters Students and 2 researchers and we are looking for postdoctoral and Ph.D. students.

Our research activities are primarily focused on developing signal processing and machine learning tools for high-resolution high-sensitivity image reconstruction from real world bio-medical imaging systems. We cover extensive medical imaging modalities such as MRI, X-ray CT, PET, and ultrasound, where one of the most important and challenging issues is overcoming the fundamental limitations of resolution and sensitivity with minimal invasiveness. Such problems pose challenges that often lead to investigations of fundamental problems in physics, mathematics, signal processing, biology, and medicine. We are unique in that we believe that these problems are an endless inspiration for new ideas and are eager to solve both specific applications and application-inspired fundamental theoretical problems in bio-medical image reconstruction.

So far, our achievements include the first demonstrations of high resolution compressed sensing dynamic MRI (k-t FOCUSS), self-reference quantitative phase microscopy, sparse dictionary learning for fMRI connectivity analysis, deep learning for low-dose CT reconstruction, non-iterative exact inverse scattering methods for diffuse optical tomography, electric impedance tomography, and elastic wave imaging. We also first discovered the mathematical link between array signal processing and multiple measurement vector problems (Compressive-MUSIC), the link between Bedrosian identity and interior tomography problem, the link between structured matrix completion and the sampling theory of finite rate of innovation (ALOHA), and the link between deep learning and convolutional framelets (Deep Convolutional Framelets). We also developed very popular toolboxes for functional near-infrared spectroscopy (NIRS-SPM) and super resolution microscopy (FALCON).

We believe in the importance of theory inspired by real world problems. An example of our theoretical work include: the inverse scattering application-inspired fundamental theory of joint sparse recovery, the accelerated MRI inspired fundamental theory of compressive sampling using low-rank interpolation, and the deep learning-inspired deep convolutional framelets theory, all of which were inspired by actual bio-medical imaging applications.

We also believe in broad, truly interdisciplinary research spanning neuroscience, biology, optics, physics, and fundamental mathematics. A strong background in these various fields is crucial in our research and we have also collaborated with medical doctors, biologists, physicists, mathematicians, and engineers to broaden our scope even further. We welcome other researchers and students from diverse backgrounds, both in academia and in industry, who wish to work with us in furthering the bounds of our understanding.