About me

My name is Jianxu Chen (pronounciation). I got my Ph.D. degree in Computer Science at the University of Notre Dame (Advisor: Danny Chen). My research area is medical image analysis. I received my B.S. degree from School of Mathematics at the University of Science and Technology of China (USTC) at 2011.

• Email: jchen16@nd.edu
• Curricula Vitae

About this website

Medical imaging is so fascinating that I would like to devote my career on solving all the variety of chanllenging problems in this field. Medical imaging is closely related to a lot of fields, such as computer vision, machine learning, statistics, and even computational geometry.

One one hand, theories, algorithms, or methodologies in different fields could find important applications in medical imaging. For instance, alpha shape (a classic concept in computational geometry) can be used to segment cells in fibrin networks (link). Another famous example is level set, a widely studied problem in numerical partial differential equations, which obtained tremendous success in medical image segmentation (link).

On the other hand, medical imaging problems are far beyond simple applications. For example, object tracking is a common problem in computer vision. However, algorithms for general object tracking in natural scene images cannot be simply applied in tracking bio-medical targets, like cells or bacteria. One of my early works was to develop a new cell tracking algorithm taking inspirations from computer vision and graph theory (link).

Recently, deep learning has penetrated into almost every hi-tech fields, including healthcare and medical imaging. The application of deep learning in medical imaging is not simply migrating models or approaches in general computer vision or machine learning to solve medical imaging problems. For example,

• Domain knowledge from human experts, like pathologists or biologists, could be much more important than those in other fields. How to incoporate human in the loop instead of playing everything in a black box is a question.
• Semi-supervised deep learning would be the next breakthrough in medical imaging. Either supervised or unsupervised learning could be problematic in this field. Obtaining sufficient high-quality training data can be troublesome in medical applications. Yet, unsupervised learning may under-esimate the importance of human supervision. Somewhere in-between could be the road for researches in the next decade.
• … … There are many more questions in deep learning specially arose in medical imaging. Feeling excited?

All in all, this site hosts all my previous works (projects, publications, source code release, etc.), current works (reading notes and papers to read), and pieces of knowledge (blog posts) related to my research in medical imaging. Hope you can find this site helpful. Enjoy!