CV
Haoran Lai
Summary
Ph.D. candidate in Biomedical Engineering at USTC. Research on 3D medical vision-language models, cross-modal alignment, and reinforcement learning for clinical reasoning.
Education
- Biomedical EngineeringUniversity of Science and Technology of China (USTC)
- Biomedical Engineering2022-06-01Southern Medical University
- Biomedical Engineering2019-06-01Southern Medical University
Work Experience
- Research Algorithm Engineer (Intern)2023-04-01 - 2025-04-01iFLYTEK Co., Ltd. — Smart Healthcare InstituteParticipated in multimodal medical AI research and development.
- XiaoYi Medical QA System: Developed medical dialogue generation module; introduced knowledge graphs and RAG techniques to improve professional accuracy.
- iFLYTEK Spark Medical LLM: Led image-text alignment algorithm research; proposed cross-modal alignment optimization module for dynamic fusion of medical image and report features.
Skills
Programming Languages
- Python
- C++
Deep Learning Frameworks
- PyTorch
- TensorFlow
Research Areas
- Medical Image Analysis
- Vision-Language Models
- Cross-modal Alignment
- 3D Medical Imaging
- Reinforcement Learning
Publications
- Bridged Semantic Alignment for Zero-shot 3D Medical Image Diagnosis2025IEEE Journal of Biomedical and Health Informatics1st author. CAS Q1, JCR Q1 Top, IF: 6.8. Proposed explicit and implicit alignment strategies for zero-shot 3D CT diagnosis.
- CARZero: Cross-Attention Alignment for Radiology Zero-Shot Classification2024CVPR 20241st author. CCF A. Proposed a cross-attention based learnable alignment strategy replacing cosine similarity for medical zero-shot classification.
- Long-Tailed Multi-Label Classification with Noisy Label of Thoracic Diseases from Chest X-Ray2024
- Prior Knowledge-Aware Fusion Network for Prediction of Macrovascular Invasion in Hepatocellular Carcinoma2022
- 2D-3D Cascade Network for Glioma Segmentation in Multisequence MRI Images Using Multiscale Information2022Computer Methods and Programs in BiomedicineCo-first author. CAS Q2, JCR Q1 Top, IF: 6.1.
- Multi-task Deep Learning Network to Predict Future Macrovascular Invasion in Hepatocellular Carcinoma2021
- Deep-Gated Recurrent Unit and Diet Network-Based Genome-Wide Association Analysis for Detecting the Biomarkers of Alzheimer's Disease2021
- Semi-supervised hybrid spine network for segmentation of spine MR images2023Computerized Medical Imaging and GraphicsCo-corresponding author. CAS Q2, JCR Q1 Top, IF: 5.4.
Presentations
- CARZero: Cross-Attention Alignment for Radiology Zero-Shot Classification2024IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2024)Seattle, WA, USAPoster presentation.
Languages
- ChineseNative
- EnglishProficient (CET-6: 488)
Interests
- 3D Medical Vision-Language Models
- Cross-Modal Feature Alignment
- Reinforcement Learning for Clinical Reasoning