[Github]
[CV]

E-mail:

Kaiwen Li

Hi, I am currently a second-year PhD student at Medical Intelligence Laboratory (MILab), Peking University, under the supervision of Assistant Professor Yanye Lu. At MILab, my research focuses on the development of multimodal large language models, exploring their applications in medicine, and weak supervision for computer vision tasks. I received my master’s degree from University of Electronic Science and Technology of China, advised by Professor Zhao Wang at the Biomedical Imaging and Vision (BIV) Lab. My research aimed to leverage artificial intelligence to analyze clinical images and data for improved diagnostic and therapeutic decision, as well as to design and implement novel medical imaging systems.

My research interests include but are not limited to:

Points-supervised fundus vessel segmentation via shape priors and contrastive learning
Kaiwen Li, Hangzhou He, Shuang Zeng, Xinliang Zhang, Yuanwei Li, Lei Zhu, Yanye Lu,
IEEE Transactions on Medical Imaging,44(12),5038-5050,2025

Seeking Interpretable Keypoint Detection and Discriminative Description for Cross-modal Fundus Image Registration
Kaiwen Li, Hangzhou He, Shuang Zeng, Ourui Fu, Xinliang Zhang, Yuanwei Li, Lei Zhu, Yanye Lu,
IEEE Transactions on Medical Imaging (Submitted)

Bridging Information Asymmetry: A Hierarchical Framework for Deterministic Blind Face Restoration
Zhengjian Yao, Jiakui Hu, Kaiwen Li, Hangzhou He, Xinliang Zhang, Shuang Zeng, Lei Zhu, Yanye Lu,
IEEE Transactions on Pattern Analysis and Machine Intelligence

Inter- and Intra-image Refinement in Few Shot Segmentation
Ourui Fu, Hangzhou He, Kaiwen Li, Xinliang Zhang, Lei Zhu, Shuang Zeng, Zhaoheng Xie, Yanye Lu,
IEEE Transactions on Image Processing (Submitted)

Point-supervised Nuclei Segmentation with Prompt-guided Superpixel
Yuanwei Li, Xinliang Zhang, Shuang Zeng, Hangzhou He, Kaiwen Li, Jiakui Hu, Lei Zhu, Yanye Lu,
IEEE Transactions on Image Processing (Submitted)

Geometry-as-context: Modulating Explicit 3D in Scene-consistent Video Generation to Geometry Context
Jiakui Hu, Jialun Liu, Liying Yang, Xinliang Zhang, Kaiwen Li, Shuang Zeng, Yuanwei Li, Haibin Huang, Chi Zhang, Yanye Lu,
CVPR2026

Chat-CBM: Towards Interactive Concept Bottleneck Models with Frozen Large Language Models
Hangzhou He, Lei Zhu, Kaiwen Li, Xinliang Zhang, Jiakui Hu, Ourui Fu, Zhengjian Yao, Yanye Lu,
ICLR2026 (Submitted)

Improve retinal artery/vein classification via channel coupling
Shuang Zeng, Chee Hong Lee, Kaiwen Li, Boxu Xie, Ourui Fu, Hangzhou He, Lei Zhu, Yanye Lu,
Expert Systems with Applications,130795,2025

Training-Free Test-Time Improvement for Explainable Medical Image Classification
Hangzhou He, Jiachen Tang, Lei Zhu, Kaiwen Li, Yanye Lu,
MICCAI2025,Spotlight

Investigation of Short-term Crystalline Lens Positional Shifts Following ICL Implantation Using SS-OCT Integrated With Multi-task Deep Learning
Zhengwei Zhang, Kaiwen Li, Zhao Wang, Jinhan Yao, Shuimiao Chang, Jiewei Liu, Xiaogang Wang,
Journal of Refractive Surgery,41(6), e569-e574, 2025

IOL tilt and decentration: a comparison of different haptic designs using CNN and SS-OCT in a short term
Jinhan Yao, Kaiwen Li, Feiyan Chai, Jingjing Wang, Zhao Wang, Xiaogang Wang,
Frontiers in Medicine, Accepted

Visible light optical coherence tomography for assessing retinal structure and function in diseased mouse models
Qianyue Gu, Wenbo Xiu, Songzhi Wu, Kaiwen Li, Baihan Li, Zhenfei Tang, Chenlu Yan, Xiaogang Wang, Chong He, Fang Lu, Zhao Wang,
Biomedical Optics Express,16(5),2077-2092,2025

Comprehensive Assessment of Anterior Segment in Refraction Corrected OCT Based on Multitask Learning
Kaiwen Li, Guangqian Yang, Shuimiao Chang, Jinhan Yao, Chong He, Fang Lu, Xiaogang Wang, Zhao Wang
Biomedical Optics Express,14(8),3968-3987,2023
[Project]

Multi-Dimensional Features Fusion of Heterogeneous Data for Predicting the Frequencies of Drug Side Effects
Kaiwen Li, Chunchun Wang, Li Zhang, Xing Chen
Briefings in Bioinformatics (Under review)
[Project]

Automatic measurement of anterior chamber angle parameters in AS-OCT images using deep learning
Guangqian Yang, Kaiwen Li, Jinhan Yao, Shuimiao Chang, Chong He, Fang Lu, Xiaogang Wang, Zhao Wang
Biomedical Optics Express,14(4),1378-1392,2023
[Project]

Comprehensive Assessment of Coronary Calcification in Intravascular OCT Using a Spatial-Temporal Encoder-Decoder Network
Chao Li, Haibo Jia, Jinwei Tian, Chong He, Fang Lu, Kaiwen Li, Yubin Gong, Sining Hu, Bo Yu, Zhao Wang
IEEE TRANSACTIONS ON MEDICAL IMAGING,41(4),857-868,2021

NSCGRN: a network structure control method for gene regulatory network inference
Wei Liu, Xingen Sun, Li Yang, Kaiwen Li, Yu Yang, Xiangzheng Fu
Briefings in Bioinformatics,23(5),2022

Coronary artery calcification and cardiovascular outcome as assessed by intravascular OCT and artificial intelligence
Jinwei Tian, Chao Li, Zhifeng Qin, Yanwen Zhang, Qinglu Xu, Yuqi Zheng, Xiangyu Meng, Peng Zhao, Kaiwen Li, Suhong Zhao, Shan Zhong, Xinyu Hou, Xiang Peng, Yuxin Yang, Yu Liu, Songzhi Wu, Yidan Wang, Xiangwen Xi, Yanan Tian, Wenbo Qu, Na Sun, Fan Wang, Yan Wang, Jie Xiong, Xiaofang Ban, Taishi Yonetsu, Rocco Vergallo, Bo Zhang, Bo Yu, Zhao Wang
Biomedical Optics Express,15(8),4438-4452,2024

  Analysis of Anterior segment-OCT (AS-OCT) images based on Multitask learning  | Using deep learning methods to automatically assess the AS-OCT images

• Built a dataset of AS-OCT images from 200 patients.
• Proposed the first multitask deep learning method for simultaneous segmentation and landmark detection in AS-OCT images, whose performance is comparable to advanced physicians.
• Our method outputs five segmentation masks including the cornea, iris, lens, ICL and IOL, and two landmark detection targets including SS and IR. Compared to previous work, we are the first to include almost all of the important structures in the anterior segment.
• Designed an automatic method to perform refraction correction for accurate quantification in AS-OCT images.
• Implemented automatic assessment of anterior segment by measuring comprehensive clinical parameters for diagnosis, surgery planning, and post-surgery assessment.

[Project]

  Prediction of the frequencies of drug side effects

• Proposed a new deep-learning network for predicting the frequencies of drug side effects by using similarity information and drug molecular structure features, achieving state-of-the-art results.
• Designed a framework for case study by performing association prediction followed by frequency prediction of drug side effects for real-world applications.
• Designed a new loss function that improves the method's ability to discriminate between discrete regression targets.

[Project]

  Robot-assisted wide-field OCT system  | Combining a robot arm with a wide-field OCT system for fast and accurate 3D imaging and modeling of objects

• Designed and built a wide-field OCT system and mounted the sample arm of the OCT system on a 6-axis robotic arm.
• Implemented the postprocessing of data and real-time display of images.
• Demonstrated system performance for in vivo imaging of entire human faces and generated human face point cloud for accurate surface reconstruction.

[Project] [Sample point cloud data of my lip]

  OCT systems for small animals  | SSOCT and SDOCT systems primarily for imaging small animals.

• Developed the software for system control, data acquisition, and preview image generation for our SSOCT and SDOCT systems.
• Implemented real-time image processing by using NVIDIA GPU and CUDA programming.
• Designed and built the data acquisition system in our SDOCT system, achieving a lateral resolution of 3.9 um.
• Participated in the design and build of the optical systems for our SSOCT and SDOCT systems.
• Conducted ex vivo imaging of mouse colon and porcine coronary arteries as well as in vivo imaging of human fingerprints, human retina, and mouse retina.

[Project]

  Multimodal endomicroscopy system  | Simultaneous OCT and fluorescence imaging based on a dual-modality catheter.

• Designed and built the control systems and data acquisition systems for OCT and fluorescence imaging.
• Developed the software for system control, data acquisition, and image generation.
• Implemented control of rotary junction and master/slave communication to perform intravascular imaging.
• Participated in the design and build of optical systems.
• Participated in the design and build of a dual-modality catheter.

[Project]