Third-year Ph.D. student · HKUST(GZ)

Hand-drawn portrait of Pengfei Hao, wearing dark glasses, a white shirt, and a navy tie.

Pengfei Hao郝鹏飞

Medical AI / Surgical Multimodal Large Language Models

I am a third-year Ph.D. student in Robotics and Autonomous Systems at HKUST(GZ), advised by Prof. Lei Zhu.

My research focuses on medical AI and surgical multimodal large language models. I am particularly interested in surgical scene understanding from images and videos, reasoning in medical multimodal large language models, and multi-agent systems.

Previously, I received my M.Phil. from HKUST(GZ) and my B.Eng. from Xi’an Jiaotong University.

Surgical AI · A conceptual illustration
Illustration of a surgeon reviewing endoscopic images. On a second monitor, a small thinking AI avatar connects surgical video evidence, anatomical regions, and a knowledge graph.
An illustration of Surgical AIFIG. 01 / CONCEPT
RESEARCH INTERESTS

Research Areas.

01

Surgical Scene Understanding

Understanding surgical images and videos, including anatomy, instruments, actions, and surgical intent.

02

Medical Multimodal Large Language Models

Integrating medical images and clinical information in multimodal large language models for medical understanding and reasoning.

01 / RESEARCH

Recent Work.

All publications
SURGICAL INTENT UNDERSTANDINGICLR 2027 submission

SurgIntent

A benchmark and knowledge-guided multi-agent framework for understanding the clinical intent behind surgical actions and the objectives they prepare for.

REASONING & REINFORCEMENT LEARNING2025
Figure 1 from the Surgery-R1 paper showing a surgical visual question, a chain of reasoning, and the Surgery-R1-54k dataset composition.

Surgery-R1

Reasoning multimodal models for surgical question answering and localization, trained with supervised fine-tuning and reinforcement learning.

MULTIMODAL SURGICAL PERCEPTIONMICCAI 2025
Surgical-MambaLLM architecture from the paper: image and question inputs, bidirectional Mamba2 integration, and localized answers.

Surgical-MambaLLM

Mamba2-enhanced cross-modal integration and instrument-aware perception for visual question localized-answering in robotic surgery.

02 / UPDATES

Recently.

Submitted , our work on surgical intent understanding, to ICLR 2027.

RESEARCH

BRAF-MLLM, our work on thyroid cancer BRAF V600E mutation prediction, has been accepted to MICCAI 2026.

CONFERENCE

Our work on knowledge-guided surgical visual reasoning appears in IEEE JBHI, Vol. 29, Issue 12.

PUBLICATION

Surgical-MambaLLM appears at MICCAI 2025.

CONFERENCE

03 / RESEARCH RECORD

Publications & manuscripts.

Published work and ongoing research

8 works

04 / BACKGROUND

Education &
background.

A background in mechanical engineering and robotics, with research interests in surgical scene understanding and medical multimodal large language models.

2024 — Present

Ph.D. in Robotics and Autonomous Systems

The Hong Kong University of Science and Technology (Guangzhou)

Advised by Prof. Lei Zhu
2022 — 2024

M.Phil. in Robotics and Autonomous Systems

The Hong Kong University of Science and Technology (Guangzhou)

2018 — 2022

B.Eng. in Mechanical Engineering

Xi’an Jiaotong University

Outstanding Undergraduate Thesis

Beyond research, I enjoy singing, football, basketball, and badminton. I previously served as president of the Feeling Music Club at Xi’an Jiaotong University.

OUTSIDE THE LAB

LET’S CONNECT

Always happy to talk research.

Surgical scene understanding and medical multimodal large language models.

phao467@connect.hkust-gz.edu.cn