Dr. Sean Zhou, 42, specializes in computer vision and data mining for medical imaging applications. Zhou, an inventor from Siemens Healthcare in Malvern, Pennsylvania, in the U.S., has developed imaging system software that employs statistical methods to identify anatomical landmarks and structures—e.g., precisely defined points in a patient’s body. This innovation is intended to make Siemens medical imaging scanners smarter, quicker, more precise, and more consistent. The technology can also be employed to improve the analysis and comparison of medical images for radiologists at the reading time.

Zhou is responsible for a series of inventions with the goal to make clinicians’ work easier by providing Automatic Landmarking and Parsing of Human Anatomy (ALPHA). ALPHA’s algorithms can automatically recognize many anatomical structures in different imaging modalities like CT, MRI, X-Ray or PET/CT, despite the fact that they look different in different patients and can also change in appearance due to imaging parameters, injuries, illnesses, or imaging artifacts.

On MRI and CT scanners, ALPHA can help medical technicians quickly plan the high-resolution scans or 3D reconstructions with the best 3D tilt and/or field of view, for the body parts to examine like head, spine or knee, i.e. where manual adjustments would be very time-consuming. The software also ensures reproducibility: “This means that a follow-up scan can be directly compared to an initial scan in order to, for example, assess the effect of treatment,” Dr. Zhou explains. For advanced multimodality image reading systems, ALPHA can help a radiologist by providing anatomical landmarks and spine labels to speed up image comparison and reporting.

Zhou was born in the Chinese province of Hubei and attended Tsinghua University, a top University in China. Having a broad interest, Zhou pursued both Engineering and Economics topics at the University and earned degrees in both fields. While studying there, Zhou also took to heart an ancient Chinese proverb carved in a stony sundial in front of the Quad: “Action speaks louder than words”, it reads. This has become a guiding inspiration for him ever since, especially in his professional life.

After moving to the United States, he worked on his Ph.D. at the University of Illinois at Urbana Champaign, completing his degree in 2002. In 2002, Zhou started his Siemens career with Corporate Research in Princeton, New Jersey. Dr. Bogdan Georgescu, who was also selected Inventor of the Year in 2012, joined him in 2004. The two scientists remain friends to this day. They worked together to develop new algorithms for automatic analysis of cardiac images. In 2005, while Georgescu continued working on the heart, Zhou moved from Corporate Research to Siemens in Malvern, Pennsylvania, and started building the ALPHA framework in a “whole-body research program”.

To design the ALPHA platform, Zhou studied the human visual recognition system as a source of inspiration. Human foveal vision system can only focus on one point at a time. However, human eyes always quickly collect a large number of evidences from an entire scene, and use the redundancy and relationships among these evidences to arrive at a reliable recognition of the scene. “ALPHA does exactly the same,” Zhou explains, “it collects a redundant set of image information to infer a target, and it also checks and confirms anatomical relationships among multiple targets.” As a result, the software is capable of achieving very high reliability, accuracy and reproducibility.

The training phase of an ALPHA algorithm is also done in a way analogous to human visual learning, i.e., purely based on examples and without explicit assumptions regarding the target anatomy or modality. As a result, the software is highly scalable to different anatomical structures, and to different imaging modalities such as CT, MRI, X-Ray, etc.
A unique capability of ALPHA is its high reliability or “self-awareness”, i.e., when the image quality is poor, or when the target is out of the field of view, or when the algorithm is failing, the software will alert the user. “This is very important for medical applications: 99% success rate alone may not be good enough if the 1% failure is difficult to catch or verify by the user.” Zhou said, “Therefore algorithms must have a high level of ‘alertness’ in order to gain enough trust from the clinicians to truly improve their productivity!”

During his ten years at Siemens, Zhou has already registered 45 individual patents in 105 patent families. Zhou is married and has two boys. During his free time, he enjoys scuba diving, fishing, cooking, and the ancient Chinese board game called “Go.”
​