The laboratory added two more international journal papers: promoting fair and credible medical imaging AI and intelligent radiotherapy
Our lab is pleased to share three publications in leading international journals, further demonstrating the team's continued breakthroughs in medical imaging intelligence, radiotherapy planning and trusted AI, covering key clinical application scenarios such as the fairness of CT multi organ segmentation, the automatic delineation of nasopharyngeal carcinoma in radiotherapy plans, and the intelligent diagnosis of retinal diseases in infants and young children. These achievements focus on the goal of "more reliable, fairer and closer to clinical process", and promote medical AI to be available and deployable in the real world.
Recent representative achievements：
- “Demographic-Aware Deep Learning for Multi-Organ Segmentation: Mitigating Gender and Age Biases in CT Image” Medical Physics
This work addresses a growing concern in clinical AI: performance disparities across patient subgroups. By introducing a demographic-aware deep learning approach, the study aims to reduce gender- and age-related biases in CT-based multi-organ segmentation, promoting more equitable and reliable model behavior across diverse populations. - “M3SegNet: A Multi-Modal and Multi-Branch Framework for Nasopharyngeal Carcinoma Segmentation in Radiotherapy Planning” IEEE Journal of Biomedical and Health Informatics (J-BHI)
This study presents M3SegNet, a multi-modal and multi-branch framework designed for nasopharyngeal carcinoma (NPC) segmentation in radiotherapy planning. By leveraging complementary information from multiple modalities and structured branch learning, the work targets robust and clinically practical delineation to support treatment planning. - “SABPI-Net: A Structure-aware Bidirectional Proxy Interaction Network for Infantile Retinal Disease Diagnosis” IEEE Transactions on Medical Imaging (TMI)
In this work, we develop SABPI-Net, a structure-aware network that models bidirectional proxy interactions to strengthen diagnostic performance for infantile retinal diseases. The approach emphasizes anatomical structure and interaction modeling, aiming to enhance accuracy and robustness under challenging pediatric imaging conditions.
We will continue to promote the research and verification of medical imaging AI in terms of generalization, interpretation, reliability and deployment, promote intelligent algorithms to truly serve clinical decision-making and treatment process, and form a high-quality transformation from scientific research innovation to clinical value.