Be water. 居善地,心善渊,与善仁,言善信,政善治,事善能,动善时。
Lecturer (Assistant Professor) in Computer Vision, Department of Computer Science, University of Sheffield
Visiting researcher in Oxford BioMedIA Group, University of Oxford
Honorary research fellow at Imperial College London
Chen (Cherise) Chen is a Lecturer in Computer Vision, at the Department of Computer Science, University of Sheffield, a core member of Insigeno Institute and Shef.AI community. Previously, she was a post-doc at Oxford BioMedIA group, University of Oxford, and the Computing Department at Imperial College London (ICL). She was also a research scientist at HeartFlow. In 2022, she obtained her Ph.D. from the Department of Computing at Imperial College London, working closely with Prof. Daniel Rueckert and Dr. Wenjia Bai.
Word cloud generated using a collection of titles from my published works. Last update: 2023
Know more about my research on the Youtube!
My main research interest lies in the interdisciplinary area of artificial intelligence (AI) and healthcare with a particular focus on medical image analysis, e.g., medical image segmentation, focusing on building and verifying robust, data-efficient, reliable machine learning algorithms to scale up AI-powered medical image analysis in real-world applications. Below are some topics I have explored:
Adversarial data augmentation
Robust machine learning
Data-efficient learning: self-supervised learning, few-shot learning, semi-supervised learning
Multi-task and multi-modal learning
Adaptive machine learning, including unsupervised domain adaptation at training/test time
Algorithms with fairness, privacy, robustness, and interpretability in mind, including uncertainty-aware training, test time model calibration
I am extremely interested in their applications including
cardiac image analysis (segmentation, registration, and shape remodeling) with quality control (e.g. topology preserving, uncertainty measurement), follow-up motion/shape analysis, survival prediction, treatment planning
prostate image analysis together with pathological image analysis, risk prediction, treatment planning
brain image segmentation and surface reconstruction for clinical applications.
[12/01/2024] One article was accepted by IEEE TMI on synthetic data generation for retinal vessel segmentation. Congrats to Linus Kreitner!
[12/10/2023] Received IEEE TMI Gold-level Distinguished Reviewer Award (2022-2023)!
[01/08/2023] One paper working on test time adaptation has been accepted by BTSD 2023 workshop, MICCAI 2023. Congrats to Jingjie GUO!
[19/07/2023] Received one of the top 12 Outstanding Reviewer awards in MICCAI 2023!
[26/06/2023] One paper working on efficient, effective vision language pre-training [link] got accepted to MICCAI 2023. Congrats to Che LIU!
[23/06/2023] Invited to give a talk at HIT-Webinar.
[25/05/2023] I am delighted to announce that I will join the Department of Computer Science at University of Sheffield as a Lecturer in Machine Vision in early Nov. 2023. I am looking for a PhD candidate to join my team, who is passionate about AI for healthcare!
[17/05/2023] Invited to give a talk at NCT Data Science Seminar 2023 at the German Cancer Research Center (DKFZ), Germany. [link]
[21/03/2023] Invited to give a talk "Advancing deep medical image segmentation with adversarial data augmentation" at the Department of Computer Science, University of Sheffield.
[03/02/2023] Invited to give a talk "Advancing deep medical image segmentation with adversarial data augmentation" at CCVL (Computational Cognition, Vision, and Learning) research group at Johns Hopkins University, US
[02/12/2022] Invited to give a talk on adversarial data augmentation at an AI giant company: SenseTime, China
[14/11/2022] Invited as a guest lecturer to give a talk to master students at TUM on "Towards Robust AI: Advanced Data Augmentation"
[02/11/2022] Hosted a lecture on "Logistic Regression" in the GirlWhoML program sponsored by Microsoft at Imperial College London
[18/09/2022] Our team won the first place in the FeTA'22 (Fetal Tissue Annotation and Segmentation Challenge)!
[28/08/2022] Our work AdvChain was published in Medical Image Analysis! [paper] [code]
[27/08/2022] Proud to receive the certificate of IEEE TMI Gold Distinguished Reviewer.
[03/08/2022] I am giving a talk on Advancing deep medical image segmentation with adversarial data augmentation at TUM AI in Medicine lab (AIM)
Title: Advancing deep medical image segmentation with adversarial data augmentation
Abstract: Deep neural networks have been successfully applied to medical image segmentation tasks, with their great potential to accelerate clinical workflows and facilitate large-scale studies. However, the performance of these deep segmentation models can be greatly impacted by changes in the data distribution due to scanner differences, varied imaging conditions as well as population shifts. To achieve satisfactory performance at deployment, these networks generally require massive labeled data collected from various domains (e.g., hospitals, scanners), which is rarely available in practice. In this talk, I will introduce our recent works on novel adversarial data augmentation algorithms to improve model generalization ability and robustness. The talk will cover two topics: a) input space adversarial data augmentation; b) feature-space adversarial data augmentation.
[02/06/2022] Three papers got accepted by MICCAI'22
[04/05/2022] One paper got early accepted by MICCAI'22
[15/04/2022] Welcome to take our CMRxMotion challenge!
[01/04/2022] Joined HeartFlow as a part-time research scientist.
[15/03/2022] Invited talk on data-efficient, robust medical image segmentation algorithms at UCL. [Slides]
[12/2021] Passed my viva. My thesis is featured in ComputerVision News! [Online Magazine] [Thesis]
[09/2021] Our work on cooperative training and latent space data augmentation was selected to give an oral presentation at MICCAI'21 [website] [video]
[09/2021] Invited talk at Imperial-Tsinghua CAS-AI Workshop
Topic: Improving the domain generalization and robustness for neural networks for medical imaging
[08/2021] Preprint: Enhancing MR Image Segmentation with Realistic Adversarial Data Augmentation is available online. [Paper] [Code]
[06/2021] Two papers were accepted at MICCAI'21 with one oral presentation
[07/2020] Four papers were accepted at MICCAI'20 with four oral presentation
[03/2020] The review paper: Deep Learning for Cardiac Image Segmentation: A Review has been published in Frontiers. [Paper]
[10/2019] Winner of the MSCMRSeg'19 [Website]
[08/2019] One paper was accepted at STACOM'19 (Oral presentation)
[07/2019] Two papers were accepted at MICCAI'19 with one oral presentation.
[07/2018] One paper was accepted at STACOM'18
A revolutionized use of atlases for test time adaptation in the medical image segmentation domain. We also investigate two different adaptation blocks: normalization blocks and dual attention blocks for better adaptability.
Keywords: test time adaptation, image segmentation, atlas, out-of-domain
MICCAI'23, Big task small data (BTSD) workshop, Oral
A plug-in data augmentation module with adversarial style augmentation, which can improve segmenaiton network's performance by 20%+ even it was trained using single domains.
Keywords: Style augmentation, Single domain generalization, image segmentation, latent space data augmentation
MICCAI'22, Early accepted
A cooperative training framework, which consists of a dual-thinking framework and a latent space data augmentation methods for single domain generalization.
Keywords: Single domain generalization, image segmentation, latent space data augmentation
MICCAI'21, Oral presentation
An adversarial data augmentation method for training neural networks for medical image segmentation, which jointly optimises bias field parameters for smart data augmentation and network parameters.
Keywords: Image segmentation, adversarial attacks, semi-supervised learning
MICCAI'20, Oral presentation
A fully automatic method to segment cardiac structures from late-gadolinium enhanced (LGE) images without using labelled LGE data for training.
Keywords: Image style transfer, image segmentation, curriculum learning
STACOM'19, Oral presentation, Winner of MSCMRSeg'19
A novel approach which learns anatomical shape priors across different 2D standard views and leverages these priors to segment the left ventricular (LV) myocardium from short-axis MR image stacks.
Keywords: Representation learning, image segmentation, shape prior
MICCAI'19, Oral presentation.
A neural-network based segmentation method for left atrial segmentation, which explores the benefits of multi-task learning to enhance representation learning.
Keywords: Multi-task learning, image segmentation, ensemble learning
STACOM'18, Ranked 4th in LASC'18 challenge.
2023:IEEE TMI Gold-level Distinguished Reviewer (2022-2023)
2023: MICCAI 2013 Outstanding Reviewer
2022: IEEE TMI Gold-level Distinguished Reviewer (2020-2022)
2022: Winner of the Fetal Tissue Annotation and Segmentation Challenge (FeTA) Challenge
2019: Winner of the Multi-sequence Cardiac MR Segmentation Challenge 2019
2012&2013: China National Scholarships (twice) (top 0.2%): The highest level of national scholarships in China