Shanghai Sixth People's Hospital · SJTU School of Medicine

AI4Rad Lab

Department of Interventional & Diagnostic Radiology · Institute of Medical Imaging, Shanghai Jiao Tong University

We develop artificial intelligence methods for disease detection, diagnostic reasoning, quantitative analysis, and image reconstruction in clinical radiology — evaluated through multicenter cohorts and radiologist reader studies, while investigating physiological and pathological mechanisms reflected in medical imaging.

AI4Rad Lab
4 Core AI Directions Vision · MLLM · Reconstruction · Simulation
7 Clinical Domains Vascular, MSK, oncology & emergency care
41 Lab Publications Peer-reviewed journals & open benchmarks
39 Team Members Clinicians, AI researchers & graduate students

Our work examines how computational models represent medical images, integrate clinical context, restore fine anatomical detail, and connect image-derived structures with physiological function.

Vision Foundation Models

01

Generalizable visual representations adaptable across imaging modalities and organs for lesion detection, anatomical segmentation, risk stratification, and quantitative biomarkers.

3D CT / MRI Pretraining Segmentation & Triage Domain Generalization

Multimodal Large Language Models

02

Combining medical images, radiology reports, clinical history, and diagnostic criteria for structured reporting, multi-agent clinical reasoning, and reader-validated decision support.

Imaging-Grounded Reasoning Structured Reports Multi-Reader Evaluation

Image Generation & Reconstruction

03

Super-resolution, denoising, artifact restoration, and cross-modal synthesis designed to preserve diagnostically critical micro-architecture and quantitative accuracy.

Super-Resolution Low-Dose Restoration Cross-Modal Synthesis

Spatial Intelligence & Physical Simulation

04

3D anatomical modeling, registration, vascular hemodynamics, and musculoskeletal biomechanics linking structural imaging phenotypes to physiological function.

Vascular Hemodynamics 3D Shape Modeling MSK Biomechanics
Clinical Imaging Focus Areas Details in About →
  • Pan-Vascular & Hemodynamic Imaging
  • Musculoskeletal & Bone Quantification
  • Oncologic Imaging & Whole-Body Metastasis
  • Emergency & Acute Abdominal Imaging
  • Longitudinal Chronic Disease Cohorts
  • Radiology Education & Reporting Feedback
  • Emerging Imaging Technologies & Translation

Selected Research

All publications →

Foundation Models, Multimodal Reasoning & Open Benchmarks Representation & Decision Systems

Large-scale representation learning, criteria-grounded clinical reasoning, and micro-architectural super-resolution validated across multicenter cohorts and structured radiologist reader studies.

Nature Biomedical Engineering · 2026 Skeletal Foundation Model

BoneCoT: Clinician Chain-of-Thought Whole-Body Skeleton Foundation Model

Built on BoneFM, a DINOv2-style self-supervised whole-body skeletal CT backbone, BoneCoT encodes clinician-derived task dependencies to guide step-by-step reasoning from anatomical localization and bone density changes to interpretable bone metastasis assessment across multicenter cohorts.

BoneFM Pretrained Backbone Clinician Task-Chain Multicenter Validation
Nature Communications · 2026 Emergency CT Triage

AbdomenNet: Acute Abdomen Diagnosis, Stratification & Triage on Non-Contrast CT

Pretrained via self-supervision on 103,989 non-contrast abdominal CT scans and fine-tuned on 5,816 annotated cases across 11 acute abdominal conditions and 3 risk-stratification tasks. Across 2,528 external patients from 3 cohorts, macro-AUROC reached 0.919 on emergent conditions; in a crossover reader study, AI assistance raised mean AUROC from 0.812 to 0.924 and cut median reading time by 52.5 s per case.

103,989 Pretraining CTs 3 External Cohorts Reader Crossover Study
npj Digital Medicine · 2026 Multimodal Agentic MLLM

Imaging-Grounded Multi-Agent MLLM for Acute Stroke Treatment

Emulates a multidisciplinary stroke team via specialized agents coordinating emergency triage, time-window estimation, multimodal neuroimaging interpretation, contraindication screening, and retrieval-augmented guideline reasoning. Evaluated on 1,658 suspected acute ischemic stroke cases across two centers, improving treatment recommendation accuracy from 0.636 to 0.785 (Macro-F1 0.562 → 0.768).

1,658 Two-Center Cases Contraindication Screening RAG Guideline Reasoning
npj Digital Medicine · 2026 Multicenter Report Benchmark

BrainMIND: Multicenter Benchmark & Reader Study on Brain MRI Reporting

Establishes a reproducible multicenter benchmark and blinded radiologist reader study evaluating 10 mainstream large language models on generating diagnostic impressions from free-text brain MRI report findings, systematically quantifying factual accuracy, clinical completeness, and reporting utility.

10 Mainstream LLMs Multicenter Brain MRI Reports Blinded Radiologist Rating
Scientific Data · 2026 Open Paired Dataset

LumbarSR: Paired Clinical CT & High-Resolution Micro-CT Vertebra Dataset

Releases the first spatially registered paired dataset of routine clinical CT and high-resolution ex-vivo micro-CT volumes of human lumbar vertebrae, bridging in-vivo clinical imaging and trabecular bone micro-architecture through an open super-resolution benchmark and community challenge.

Clinical CT ↔ Micro-CT Trabecular Micro-Structure Open Benchmark & Challenge
J. Med. Internet Res. · 2026 Emergency Clinical Reasoning

HR-LLM-Stroke: Hierarchical Multi-Agent Emergency Stroke Decision Support

Addresses rapid multi-source clinical synthesis within narrow stroke reperfusion windows by orchestrating specialized agents for structured imaging readout, neurological feature extraction, and guideline-aligned treatment recommendation, validated across multicenter cohorts with an interactive web demo.

Time-Critical Stroke Triage Multicenter Validation Open Code & Live Demo

Hemodynamic Modeling, Quantitative Biomarkers & Clinical Studies Biophysics & Quantitative Imaging

Combining 3D vascular geometry, hemodynamic constraints, biomechanical trajectory simulation, and multi-reader clinical studies across vascular, cardiopulmonary, and musculoskeletal disorders.

IEEE J. Biomed. Health Inform. · 2026 Vascular Hemodynamics · Physics-Guided DL

FlowPredictor–Corrector: Hemodynamics-Guided Two-Step Vascular Flow Prediction

Integrates 3D vascular point-cloud representation learning with physics-based hemodynamic correction to non-invasively map 3D vascular geometry to 4D blood-flow fields, preserving physical consistency with reference CFD simulations and clinical stenosis evaluations at rapid inference speed.

3D Point-Cloud Geometry Physics-Based Correction Rapid 4D Flow Prediction
Insights into Imaging · 2026 Vascular MRI Reporting · Multi-Reader Study

Criteria-Guided LLM Assistance for Carotid Plaque AHA Classification

Evaluated 10 LLMs and a 4-radiologist reader study across 433 multicenter patients with free-text carotid MRI reports, demonstrating that explicitly encoding modified-AHA staging criteria improves inter-reader agreement and enables effective deployment of a lightweight 8B model.

433 Multicenter Patients Explicit Staging Criteria Local 8B Model + 4 Readers
Academic Radiology · 2026 Cardiopulmonary CTA · Imaging Biomarker

Pulmonary Artery-to-Vein Volume Difference in Acute Pulmonary Embolism

Grounded in cardiopulmonary hemodynamics — elevated pulmonary vascular resistance and reduced left-atrial venous return — this study establishes 3D pulmonary artery-to-vein volume difference on chest CTA as an interpretable quantitative biomarker for acute pulmonary embolism risk stratification.

Cardiopulmonary Physiology 3D CTA Volumetry Acute PE Risk Stratification
J. Orthop. Surg. Res. · 2026 Cervical Anatomy · 3D Surgical Simulation

3D CTA Trajectory Simulation for C1 Screw Placement in Ponticulus Posticus

Quantifies vertebral artery injury risks associated with ponticulus posticus anatomical variants via 3D pedicle screw trajectory simulation across 141 cervical CTA examinations, establishing a preoperative anatomical screening workflow.

141 Cervical CTA Subjects 3D Screw Trajectory Simulation Preoperative Risk Quantification

Latest News

Browse archive →

Loading the latest news…

Join Us

  • 100+ GPU Dedicated Cluster (>100 kW in-house compute)
  • PB-Scale Dedicated Storage (multimodal imaging cohorts)
  • Well-Funded (>10 national grants · >¥10M RMB in past 5 years)
Platform details →
Clinical–Engineering Integration

Interdisciplinary Research Group

Led by Prof. Yuehua Li together with PIs Ruipeng Zhang, Jianyong Wei, and Zheng Sun, our group brings together radiologists, engineering researchers, postdoctoral fellows, and graduate students inside a high-volume clinical imaging department.

Recruitment & Collaboration

Work With Us

We welcome multicenter collaborations and recruit PhD and master’s students, postdoctoral fellows, and visiting researchers in medical imaging AI, computer vision, and multimodal learning. Motivated undergraduates from SJTU School of Medicine are also encouraged to participate in RBL (Research-Based Learning) and innovation competitions. Please contact [email protected].