Curriculum Vitae

Education

• Ph.D. in Biomedical Engineering, University of Science and Technology of China (USTC), 2022 – Present
  Advisor: Prof. Shaohua Zhou (IEEE Fellow, Dean)

• M.S. in Biomedical Engineering, Southern Medical University, 2019 – 2022
  Advisor: Prof. Meiyan Huang (Guangdong Young Pearl River Scholar)

• B.S. in Biomedical Engineering, Southern Medical University, 2015 – 2019

Research Interests

Multimodal learning, medical image analysis, vision-language models, cross-modal alignment, 3D medical imaging, reinforcement learning for clinical reasoning.

Internship Experience

iFLYTEK Co., Ltd. — Smart Healthcare Institute
Research Algorithm Engineer (Intern), Apr 2023 – Apr 2025

• XiaoYi Medical QA System: Developed and optimized the medical dialogue generation module for a multimodal medical QA system. Introduced knowledge graphs and RAG techniques to significantly improve professional accuracy and consistency in clinical dialogue, supporting applications in medical record QA, imaging-assisted diagnosis, and health consultation.
• iFLYTEK Spark Medical LLM: Led research on image-text alignment algorithms and multimodal pre-training strategies. Proposed a cross-modal alignment optimization module for dynamic fusion of medical image and report features, achieving notable performance gains on multimodal diagnosis, report generation, and clinical QA tasks.

Research Experience

3D Medical Vision-Language Models, Jul 2024 – Present

• Explored construction of 3D medical VLMs for automated understanding and interpretable reasoning over complex 3D medical scenes.
• Built a VQA dataset from 3D CT scans and paired reports; proposed an efficient image-text alignment strategy; designed a loss correction strategy to address output bias in autoregressive models caused by report spatial bias.
• In the RL stage, proposed a consistency reward function to optimize reasoning chains, significantly enhancing diagnostic accuracy and logical coherence.
• The resulting 3D CT LLM surpasses existing open-source general and medical models on diagnostic and reasoning benchmarks.
• Paper submitted to CVPR 2026 (CCF A).

3D Vision-Language Alignment and Cross-Modal Reasoning, Dec 2023 – Jun 2024

• Designed cross-modal alignment algorithms for semantic mapping between 3D vision and language, supporting zero-shot classification and image-text retrieval on 3D medical images.
• Proposed a combined explicit and implicit alignment strategy to address limited 3D CT-report data and suboptimal alignment; achieved state-of-the-art (SOTA) performance on three open-source 3D CT benchmarks.
• Published in IEEE Journal of Biomedical and Health Informatics (CAS Q1, JCR Q1 Top, IF: 6.8).

2D Vision-Language Alignment and Cross-Modal Reasoning, Sep 2022 – Nov 2023

• Built a zero-shot image classification and recognition framework based on self-supervised contrastive learning and deep feature modeling.
• Proposed a learnable cross-attention based image-text alignment strategy to replace manually defined metrics (e.g., cosine similarity); achieved SOTA performance on chest X-ray zero-shot diagnosis and object detection tasks.
• Published at CVPR 2024 (CCF A).

Early Macrovascular Invasion Prediction via Deep Learning, May 2020 – Apr 2022

• Developed multimodal diagnostic models for predicting early macrovascular invasion in hepatocellular carcinoma patients, assisting clinical treatment decisions.
• Built a multimodal model from liver CT and clinical indicators; introduced word vector encoding for clinical feature representation; further proposed a CT-only prediction framework incorporating graph convolutional networks for tumor boundary and global liver structure features.
• Published in IEEE Transactions on Medical Imaging (CAS Q1, IF: 10.048) and EClinicalMedicine (CAS Q1, IF: 17.033).

Alzheimer’s Disease Imaging-Genomics Cross-modal Modeling, Sep 2019 – May 2020

• Explored cross-modal association patterns between medical images and gene sequences; built a unified multimodal representation learning framework for Alzheimer’s disease (AD).
• Proposed a feature alignment method for brain MRI and SNP sequences; identified key SNP loci linked to brain structural changes via gradient backpropagation to reveal potential AD-related pathogenic mechanisms.
• Published in Medical Image Analysis (CAS Q1, JCR Q1 Top, IF: 10.7).

Publications

Journal Papers

1. Bridged Semantic Alignment for Zero-shot 3D Medical Image Diagnosis
   Haoran Lai, et al. IEEE Journal of Biomedical and Health Informatics, 2025. (1st author) — CAS Q1, JCR Q1 Top, IF: 6.8

2. Prior Knowledge-Aware Fusion Network for Prediction of Macrovascular Invasion in Hepatocellular Carcinoma
   Haoran Lai, et al. IEEE Transactions on Medical Imaging, 2022. (1st author) — CAS Q1, JCR Q1 Top, IF: 10.048

3. Multi-task Deep Learning Network to Predict Future Macrovascular Invasion in Hepatocellular Carcinoma
   Haoran Lai*, et al. EClinicalMedicine, 2021. (co-first author) — CAS Q1, JCR Q1 Top, IF: 17.033

4. Deep-Gated Recurrent Unit and Diet Network-Based Genome-Wide Association Analysis for Detecting the Biomarkers of Alzheimer’s Disease
   Medical Image Analysis, 2021. (2nd author) — CAS Q1, JCR Q1 Top, IF: 10.7

5. 2D-3D Cascade Network for Glioma Segmentation in Multisequence MRI Images Using Multiscale Information
   Haoran Lai*, et al. Computer Methods and Programs in Biomedicine, 2022. (co-first author) — CAS Q2, JCR Q1 Top, IF: 6.1

6. Semi-supervised hybrid spine network for segmentation of spine MR images
   Computerized Medical Imaging and Graphics, 2023. (co-corresponding author) — CAS Q2, JCR Q1 Top, IF: 5.4

Conference Papers

1. CARZero: Cross-Attention Alignment for Radiology Zero-Shot Classification
   Haoran Lai, et al. CVPR 2024. (1st author) — CCF A, Top Venue in Computer Vision

2. Long-Tailed Multi-Label Classification with Noisy Label of Thoracic Diseases from Chest X-Ray
   Haoran Lai, et al. ISBI 2024. (1st author) — IEEE Flagship Conference in Biomedical Imaging

Under Review

1. E3D-GPT: Enhanced 3D Visual Foundation for Medical Vision-Language Model
   Haoran Lai, et al. (1st author)

2. Med3D-R1: Incentivizing Clinical Reasoning in 3D Medical Vision-Language Models for Abnormality Diagnosis
   Haoran Lai, et al. MICCAI 2026 (under review). (1st author) — CCF B

Conference Presentations

• CVPR 2024, Seattle, WA, USA — Poster presentation on CARZero, Jun 2024

Honors & Awards

• Outstanding Ph.D. Graduate, USTC, 2026
• National Scholarship for Ph.D. Students, USTC, 2024
• Suzhou Industrial Park Scholarship, USTC, 2022
• Outstanding Graduate Student, Southern Medical University, 2022
• National Scholarship for M.S. Students, Southern Medical University, 2021
• Exempt Admission Scholarship (推免生奖学金), Southern Medical University, 2019
• Champion, MICCAI Thyroid Nodule Classification & Segmentation (TN-SCUI2020) — Classification Track, 2020

Skills

• Programming: Python (proficient), C++ (proficient)
• Frameworks: PyTorch, TensorFlow
• Languages: Chinese (native), English (CET-6: 488)
• Certifications: National Computer Rank Examination Level 2, C++ (98/100)