my photo

Short Bio

I am now an associate professor and master supervisor at the School of Intelligent Systems Engineering, Sun Yat-Sen University.
I received my Ph.D. and bachelor's degrees from Department of Electrical Engineering in Tsinghua University in 2010 and 2015, respectively. Then, I worked as an assistant professor in Shenzhen Insittute of Advanced Techonology from 2015 to 2018. After this, I worked as a post-doctoral researcher in Digital Imaging Group of Western University with Prof. Shuo Li from 2018 to 2020.
I am now mainly focusing on in-depth research of the frontier problems in medical image analysis by artificial intelligence. Also, I have the acadamic background of electrical engineering and plasma science. If you are interested in artificial intelligence in medical applications, electrical engineering, and plasma science, please contact me.
Let us explore the unknown scientific world together.


Research Interests

Medical Image Analysis, Artificial Intelligence, Deep Learning, Computer Vision.

My research mainly aims at the in-depth research of medical image analysis and artificial intelligence, for human tissue detection/segmentation, disease diagnosis, and treatment planning. The methodologies that I am interested in are: object detection, graph reasoning, and weakly-supervised learning architectures.
I recently investigate graph reasoning in tissue detection and disease diagnosis tasks for weakly labelled data, for promoting more powerful and less dependent AI-powered healthcare.
I am now looking for Masters/RA/Postdoc to work on medical image analysis, especially on graph reasoning and weakly-supervised learning architectures.


Upcoming Events

  • I'm serving as a reviewer for MICCAI 2020.
  • My paper in MICCAI 2019 has been nominated for the Young Scientist Award and are invited to a special issue in MEDIA (Top journal on Medical image analysis with Impact Factor: 8.88). .

News

  • [10/2020] One paper accepted to MEDIA.
  • [06/2020] Two papers accepted to MICCAI.
  • [07/2019] One paper accepted to MEDIA.
  • [07/2019] One paper accepted to MICCAI with early acceptance and Young Scientist Award nomination.
  • [05/2019] Our paper published in JBHI are nominated as frequently cited papers ESI.

Selected Publications

For a complete list, please check my google scholar.
Codes for these papers will be uploaded to github in succession in the near future. Also, I am very happy to provide my codes in advance if you are interested in my papers for academic use.

【04】 Discriminative dictionary-embedded network for comprehensive vertebrae tumor diagnosis.
Shen Zhao , Bin Chen, Heyou Chang, Xi Wu, and Shuo Li.
Medical Image Computing and Computer Assisted Intervention (MICCAI) 2020. Discriminative dictionary-embedded network (DECIDE) .

  • Methodoloy innovations: Based on our previous work in MEDIA 2020 and MICCAI 2019, we go further into dictionary learning in recognition task by using projection-guided dictionary learning; also, we advance the recognition task to tumor diagnosis, a crucial clinical application for treatment planning and metastasis preventation, by designing a label-consistent dictionary learning network.
    This work could benefit other object recognition problems where the locations of the target objects are sparse in the input image.
  • Advantages: An heuristic exploration of leveraging sparse codes to encode different recognized objects (vertebrae) for better distinguishability.
    A compact end-to-end network for simultaneous vertebrae recognition and tumor diagnosis.
  • Disadvantages: Not found yet.

【03】 Automatic Vertebrae Recognition from Arbitrary Spine MRI images by a Category-Consistent Self-calibration Detection Framework.
Shen Zhao , Xi Wu, Bo Chen, and Shuo Li.
Medical image analysis (MEDIA) 2020. The overall workflow and the detailed modules of the proposed network (Can-See).

  • Methodoloy innovations: Based on our previous work published in MICCAI 2019, we go deeper to the principles of message passing to present a more profound understanding of how the label compatibility matrix is trained and how it is used in the testing phase.
    We also provide a preliminary attempt to use dictionary learning in deep CNN's by proposing a label-consistent dictionary learning module and integrating it into the deep learning-based pre-recognition network.
  • Advantages: An in-depth understanding of message passing, dictionary learning, as well as the RPN networks for object detection.
  • Disadvantages: The dictionary learning method is only a preliminary attempt, and the k-sparse auto-encoder is somewhat simple.

【02】 Automatic Vertebrae Recognition from Arbitrary Spine MRI Images by a Hierarchical Self-calibration Detection Framework.
Shen Zhao , Xi Wu, Bo Chen, and Shuo Li.
Medical Image Computing and Computer Assisted Intervention (MICCAI) 2019. The self-calibration recognition network (SRN).

  • Methodoloy innovations: For the first time, message passing is formulated into deep learning object recognition network for self-awareness and self-calibration of wrong recognition results of an object detection network.
    This work could benefit other object recognition problems where the locations of the target objects have some certain internal spatial relationships.
  • Advantages: An efficient way of leveraging the label relationships of different recognized objects (vertebrae) for errors in object recognition networks.
    An in-depth understanding of the two-stage object recognition networks.
  • Disadvantages: Not found yet.

【01】 Automatic spondylolisthesis grading from MRIs across modalities using faster adversarial recognition network.
Shen Zhao , Xi Wu, Bo Chen, and Shuo Li.
Medical image analysis (MEDIA) 2019. The multi-task recognition-diagnosis network that recognizes critical vertebrae and diagnoses spondylolisthesis gradings in an adversarial manner.

  • This is the first paper that I use deep learning-based methods for medical image analysis.
  • Methodoloy innovations: A multi-task detection-diagnosis network has been constructed for simultaneous human organ (vertebrae) recognition and disease (spondylolisthesis) diagnosis.
    Adversarial networks are attempted to be used in recognition networks. Hybrid supervision is introduced to provide stronger supervision signals to the proceeding network.
  • Advantages: Accurate recognition of critical vertebrae (which tends to suffer from spondylolisthesis), and precise diagnosis of the disease.
  • Disadvantages: As my research goes deeper, I find that the adversarial module and the auxiliary path for gradient back-propagation may be substituted by more efficient methods.

【00】 Robust Segmentation of Intima-Media Borders with Different Morphologies and Dynamics During the Cardiac Cycle.
Shen Zhao , Zhifan Gao, Heye Zhang, Yaoqin Xie, et.al.
IEEE Journal of Biomedical and Health Informatics (JBHI) 2017.
The adaptive snake algorithm for segmenting the IM borders of the next frame.

  • This is the first paper after I entered medical image analysis domain!
  • Methodoloy innovations: A robust grayscale-derivative constraint snake algorithm that avoids sudden change of grayscale and derivative at IM border points between consecutive frames.
  • Advantages: Fast and robust segmentation of IM borders in ultrasound sequences with noises, large movements between frames, different IM appearances, and plagues.
  • Potential clinical usage: Diagnosis of cardiovascular diseases not only by IMT but also by its changes during the cardiac cycle.
  • Disadvantages: Snake parameters are still based on manual experience.
  • Codes are available for this paper.


Professional Activities

Membership of IEEE, CCF.

Conference Services:

Reviewer of MICCAI'20, MICCAI'19

Journal Reviews:

Medical Image Analysis
IEEE Journal of Biomedical and Health Informatics
Computerized Medical Imaging and Graphics

MOTTO

  • "Constantly working hard with self-discipline, profoundly abiding morality with social commitment." ——Tsinghua University motto.

  • "Superior horses can hardly reach ten steps in one go, however, inferior horses can run a long way provided they do not give up." ——Xun Zi, a paper for urging us to study.


  • Last updated date: Apr 2019.

    Click the map to see page view statistics since Jan 2019.