|
Education
2018 - 2023: Ph.D. in Computer Science, HIT, China.
2016 - 2018: M.Sc. in Optical enginering, HIT, China.
2010 - 2014: B.Sc. in Optical and Electrical enginering , CUST, China.
|
|
News
01/2025: Our "Convergent–diffusion denoising model for multi-scenario CT image reconstruction" is accepted by Computerized Medical Imaging and Graphics
12/2024: Our "Adjacency-aware Fuzzy Label Learning for Skin Disease Diagnosis" is accepted by IEEE Transactions on Fuzzy Systems
12/2023: Our "Synergistically Learning Class-specific Tokens for Multi-class Whole Slide Image Classification" is accepted by BIBM
09/2023: Wow! Two MedIA paper got accepted in two months! Our "Ambiguity-aware breast tumor cellularity estimation via self-ensemble label distribution learning" is accepted by MedIA
08/2023: Our paper, "Curriculum label distribution learning for imbalanced medical image segmentation" is accepted by MedIA
01/2023: We have relased the MICCAI INSTANCE 2022 challenge summarized paper on arxiv, "The state-of-the-art 3D anisotropic intracranial hemorrhage segmentation on non-contrast head CT: The INSTANCE challenge" arxiv
09/2022: The INSTANCE 2022 challenge has successfully finished! Ten out of 74 teams won the prizes. We have released the challenge results on the test set Results.
06/2022: Our paper "ULTRA: Uncertainty-Aware Label Distribution Learning for Breast Tumor Cellularity Assessment" is accepted by MICCAI 2022
04/2022: The training data of the INSTANCE 2022 challenge has beed released!
03/2022: We will host an Intracranial Hemorrhage segmentation challenge INSTANCE 2022 at MICCAI2022. Welcome to participate in it.
08/2021: Our paper "Hematoma Expansion Context Guided Intracranial Hemorrhage Segmentation and Uncertainty Estimation" is accepted by IEEE-JBHI
|
|
Research
I do research on medical image analysis and computer vision, with a focus on learning with imperfect labels. Recently, I am also working on topics including prompt leaning in medical image analysis and diffusion models.
|
|
|
Convergent–diffusion denoising model for multi-scenario CT image reconstruction
Xinghua Ma, Mingye Zou, Xinyan Fang, Gongning Luo, Wei Wang, Suyu Dong, Xiangyu Li, Kuanquan Wang, Qing Dong, Ye Tian, Shuo Li
Computerized Medical Imaging and Graphics
Paper
A new approach for CT image reconstruction by employing a novel diffusion-based framework, tailored for the CTIR task, and focusing on imaging noise with stepwise inference.
The proposed method introduces an innovative dual-domain integration scheme, incorporating sinogram semantics to guide artifact prediction and effectively avoiding secondary artifacts.
|
|
|
Adjacency-aware Fuzzy Label Learning for Skin Disease Diagnosis
Murong Zhou, Baifu Zuo, Guohua Wang, Gongning Luo, Fanding Li, Suyu Dong, Wei Wang, Kuanquan Wang, Xiangyu Li, Lifeng Xu
IEEE Transactions on Fuzzy Systems
Paper
The proposed AFLL framework is the first to address the challenge of distinguishing adjacent categories in acne severity grading tasks.
A novel way to mitigate the impact of errors in preceding decisions on subsequent ones by dynamically and selectively accessing prior predictions in a selective masking decision strategy.
|
|
|
Synergistically Learning Class-specific Tokens for Multi-class Whole Slide Image Classification
Pengzhong Sun, Wei Wang, Xiangyu Li, Suyu Dong, Shuo Li, Kuanquan Wang, Gongning Luo
IEEE International Conference on Bioinformatics and Biomedicine (BIBM)
Paper
The first work to explore applying multiple class-specific tokens to learn class-specific information for multi-class WSI analysis tasks.
A novel dynamic class-centric training strategy that integrates multiple loss functions by a dynamic weighting mechanism to ensure robust class-specific token learning.
|
|
|
Ambiguity-aware breast tumor cellularity estimation via self-ensemble label distribution learning
Xiangyu Li, Xinjie Liang, Gongning Luo, Wei Wang, Kuanquan Wang, Shuo Li
Medical image analysis (MedIA),2023
Paper| Code
The first work that simultaneously model inter-rater and intra-rater ambiguity of tumor cellularity(TC) estimation tasks.
The proposed method significantly improves both segmentation-based and regression-based methods on the TC estimation task.
|
|
|
Curriculum label distribution learning for imbalanced medical image segmentation
Xiangyu Li, Gongning Luo, Wei Wang, Kuanquan Wang, Shuo Li
Medical image analysis (MedIA),2023
Paper| Code
The first work to solve the label distribution imbalance problem in LDL-based segmentation tasks.
The proposed CLDL effectively combines curriculum learning and deep label distribution learning in a unified framework.
|
|
|
ULTRA: Uncertainty-Aware Label Distribution Learning for Breast Tumor Cellularity Assessment
Xiangyu Li, Xinjie Liang, Gongning Luo, Wei Wang, Kuanquan Wang, Shuo Li
International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI)
Paper| Code
The first work to model label ambiguity of tumor cellularity(TC) by transfferring the TC score regression to a label distribution learning problem.
The proposed method significantly improves both segmentation-based and regression-based methods on the TC estimation task.
|
|
|
Hematoma Expansion Context Guided Intracranial Hemorrhage Segmentation and Uncertainty Estimation
Xiangyu Li, Xinjie Liang, Gongning Luo, Wei Wang, Kuanquan Wang, Shuo Li
IEEE Journal of Biomedical and Health Informatics (J-BHI)
Paper| Code
A new insight for intracranial hemorrhage segmentation by incorporating the hematoma expansion
into the segmentation network.
A novel way to exploit context information in intracranial hemorrhage segmentation by directly modeling the hematoma variation between two adjacent slices.
|
|
