AI Medical Imaging in China – Large Models, AI Doctors, and Multi-Modal Diagnostics

From 100-billion-parameter foundation models to AI pediatricians serving 150+ county-level hospitals, China is integrating artificial intelligence into medical imaging at an unprecedented scale and speed.

I. Introduction: AI Transforms Medical Imaging

Artificial intelligence is transforming medical imaging at an accelerating pace in China. What was once a field dominated by single-point solutions — AI algorithms for specific tasks like detecting pulmonary nodules or diagnosing diabetic retinopathy — has evolved into a landscape of foundation models, multi-modal AI systems, and integrated clinical decision support.

In 2026 alone, multiple AI medical devices received NMPA Class III certification — the highest regulatory classification for medical devices in China — covering digital pathology, chromosome analysis, and surgical robotics.

This article reviews the most significant AI medical imaging breakthroughs in China in 2026.

I. iMedImage: The 100-Billion-Parameter Foundation Model

In 2026, Deshi Biotech (德适生物) unveiled iMedImage, a medical imaging foundation model with 100 billion parameters — making it one of the largest medical AI models in the world.

Technical Specifications

Feature	Specification
Model size	100 billion parameters
Imaging modalities supported	19 modalities (CT, MRI, X-ray, ultrasound, pathology, and more)
Clinical scenario coverage	Over 90% of clinical imaging scenarios
Training data	Hundreds of millions of medical images from diverse sources

AI AutoVision: Chromosome Karyotype Analysis

The iMedImage foundation model powers AI AutoVision, an automated chromosome karyotype analysis system that:

• Reduces analysis time — from hours to just 2.5 minutes per image
• Achieves over 99% accuracy — comparable to expert cytogeneticists
• Automates manual tasks — chromosome segmentation, classification, and karyotyping

AI AutoVision received NMPA Class III certification in May 2026, making it one of the first AI-powered chromosome analysis systems to achieve regulatory clearance in China.

II. DeepFAN: Asia's First AI-Assisted Pulmonary Nodule Differentiation Trial

In April 2026, a multicenter study led by Peking Union Medical College Hospital (PUMCH) completed Asia's first registered clinical trial of AI-assisted benign-malignant differentiation of pulmonary nodules.

The DeepFAN Model

DeepFAN is a deep learning model that:

• Analyzes chest CT scans to differentiate benign from malignant pulmonary nodules
• Achieves diagnostic accuracy comparable to that of experienced radiologists
• Reduces inter-observer variability in nodule assessment

Clinical Trial Results

The multicenter trial, involving multiple Chinese hospitals, demonstrated that DeepFAN achieved:

• Area under the ROC curve (AUC) exceeding 0.92
• Sensitivity and specificity comparable to expert radiologists
• Significant reduction in reading time for radiologists

DeepFAN received NMPA Class III medical device registration, making it one of the first AI-assisted pulmonary nodule analysis tools to achieve full regulatory clearance in China.

III. AI Pediatrician: 24 Million Records, 150+ Hospitals

In 2026, the AI Pediatrician — an AI system trained on 24 million pediatric medical records — was deployed in over 150 county-level hospitals across China.

Technology and Training

The AI Pediatrician is a large language model (LLM) trained on:

• 24 million pediatric medical records — including clinical notes, lab results, and imaging reports
• Diagnostic guidelines and treatment protocols from leading pediatric hospitals
• Continuous learning — the model is updated regularly with new data

Clinical Impact

The AI Pediatrician supports clinical decision-making in resource-limited settings by:

• Differential diagnosis — suggesting possible diagnoses based on symptoms and test results
• Treatment recommendations — providing evidence-based treatment options
• Referral triage — identifying patients who need referral to higher-level facilities

In a study evaluating the system's performance, the AI Pediatrician achieved diagnostic accuracy exceeding 85% for common pediatric conditions, with a significant reduction in diagnostic errors compared to baseline.

IV. "Xiao Jun Doctor 2.0": AI for Cranial CT Diagnostics

In January 2026, Beijing Tiantan Hospital launched "Xiao Jun Doctor 2.0" — an AI-powered cranial CT diagnostic agent that:

• Covers nearly 100 diseases in cranial CT imaging
• Generates reports in under 1 minute — compared to 15–20 minutes for manual reading
• Achieves diagnostic accuracy of nearly 80% — comparable to junior to mid-level radiologists
• Deployed in 40+ medical institutions across 14 provinces and cities

The system is particularly valuable for emergency settings, where rapid triage of cranial CT findings can be life-saving.

V. Chest CT Large Model: First Multi-Disease AI in NMPA Review

In 2026, China's first large-model multi-disease AI product for chest CT imaging was accepted into the NMPA's Innovative Medical Device Special Review Process — marking a new milestone for AI medical devices in China.

The system simultaneously detects multiple abnormalities in chest CT scans, including:

• Pulmonary nodules and lung cancer
• Coronary artery calcification
• Chronic obstructive pulmonary disease (COPD)
• Osteoporosis and vertebral fractures
• Incidental findings

VI. Other AI Medical Device Approvals in 2026

In 2026 alone, multiple AI medical devices received NMPA Class III certification:

Product	Company	Application	Approval Date
AI AutoVision	Deshi Biotech	Chromosome karyotype analysis	May 2026
DeepFAN	PUMCH / Deepwise	Pulmonary nodule differentiation	April 2026
Cervical cytology AI	Yice Technology	Digital pathology screening	Early 2026
ROPA6	Changmugu Medical	Full-orthopedic surgical robot	March 2026

VII. What This Means for International Patients

For international patients considering diagnostic workup in China:

• Faster diagnosis — AI-powered imaging reduces reporting times from hours to minutes
• More accurate detection — AI systems detect subtle findings that may be missed in manual reading
• Consistent quality — AI reduces inter-reader variability
• Cost advantage — AI-powered imaging in China is significantly more affordable than in Western countries

Conclusion

AI is transforming medical imaging in China, moving from single-purpose algorithms to foundation models that support comprehensive diagnostic workflows. With the iMedImage 100-billion-parameter model, the AI Pediatrician serving 150+ county-level hospitals, and multiple AI devices achieving NMPA Class III certification, China is establishing itself as a global leader in AI-assisted medical imaging.