AI-Enhanced Medical Imaging Takes Center Stage at MBZUAI Commencement: Chao Qin's Breakthrough Tools and Largest Open-Source Ultrasound Dataset

Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), the world's first dedicated graduate research university for artificial intelligence, recently celebrated its Commencement 2026 ceremony in Abu Dhabi, spotlighting groundbreaking advancements in AI-enhanced medical imaging. Among the standout achievements was the work of PhD candidate Chao Qin from the Computer Vision program, whose innovative AI tools and the creation of the largest open-source ultrasound dataset are poised to transform disease detection and diagnosis in healthcare settings across the United Arab Emirates and beyond.

The event underscored MBZUAI's pivotal role in UAE's National AI Strategy 2031, which aims to position the country as a global leader in AI by integrating it into key sectors like healthcare. With over 149 graduates from diverse backgrounds, the ceremony featured speeches from University President Eric Xing and valedictorian Hanoona Rasheed, emphasizing ethical AI development and real-world impact. Qin's presentation highlighted how his research bridges the gap between cutting-edge AI models and practical clinical applications, drawing attention from healthcare leaders and policymakers.

Chao Qin's Journey at MBZUAI: Pioneering AI for Healthcare

Chao Qin, a PhD candidate in Computer Vision at MBZUAI, entered the field of medical imaging without prior experience but quickly pivoted his research to address critical needs in diagnostic accuracy. His motivation stemmed from the potential of AI to alleviate the burden on radiologists, who often face overwhelming workloads and variability in image quality. 'This was a completely new area for me, but MBZUAI's supportive environment allowed me to dive deep,' Qin shared during the commencement spotlight.

Under the guidance of faculty like Mohammad Yaqub, whose expertise spans AI in ultrasound, MRI, and CT imaging, Qin developed tools that adapt general-purpose AI models for specialized medical tasks. This aligns with MBZUAI's interdisciplinary approach, fostering collaborations between computer science and healthcare domains. Qin's trajectory exemplifies how UAE universities are nurturing global AI talent, with MBZUAI attracting students from over 24 countries and producing graduates ready to tackle regional challenges like rising chronic diseases.

DB-SAM: Revolutionizing Universal Medical Image Segmentation

One of Qin's flagship contributions is DB-SAM (Dual-Branch Segment Anything Model), a framework that adapts Meta's Segment Anything Model (SAM) for high-quality segmentation across diverse medical images. Traditional SAM excels in natural images but struggles with the domain gap in medical data, where textures, contrasts, and artifacts differ significantly.

DB-SAM employs a dual-branch architecture: one branch fine-tunes image encoder features for medical specificity, while the other leverages prompt-aware decoders. Evaluated on 30 public datasets encompassing 2D and 3D images (ultrasound, CT, MRI, X-rays), it achieved an 8.8% absolute gain in Dice score on 21 3D tasks compared to prior adapters. This means more precise outlining of organs, tumors, and lesions, crucial for treatment planning.The MICCAI 2024 Oral paper details these results, earning Qin Best Paper Finalist and Young Scientist Award nods.

In UAE clinics, where imaging volume is surging due to population growth, DB-SAM could reduce segmentation time from hours to seconds, enabling faster interventions. Early pilots with partners like Cleveland Clinic Abu Dhabi show promise in real-time applications.

OpenUS: The Largest Open-Source Ultrasound Dataset Unlocking AI Potential

To train robust models, Qin compiled OpenUS, the world's largest open-source ultrasound dataset with 471,000 images from 53 different scanners across 42 public sources. Ultrasound imaging, vital for non-invasive disease detection in cardiology, obstetrics, and oncology, suffers from variability in probe types, patient conditions, and operator skills.

OpenUS uses self-adaptive masked contrastive learning to pretrain foundation models, improving downstream tasks like segmentation and classification by 5-10% on benchmarks. By democratizing access, it empowers researchers worldwide, particularly in resource-limited settings. 'Diverse data ensures reliability across hospitals and machines,' Qin noted.The arXiv preprint outlines the methodology and benchmarks.

In the UAE, where ultrasound is a frontline tool for detecting conditions like breast cancer (affecting 1 in 8 women regionally), OpenUS supports national goals for precision medicine. MBZUAI's release aligns with open science initiatives, fostering global collaboration.

AI's Role in Enhancing Ultrasound Disease Detection

Ultrasound detects diseases like tumors, heart defects, and fetal anomalies with high sensitivity but operator-dependent accuracy (typically 80-90%). AI boosts this: studies show AI-assisted systems reach 94% accuracy in early cancer detection, reducing false negatives by 20%.Gulf News reports on Qin's tools flagging anomalies in seconds.

Qin's systems process videos, highlighting irregularities in real-time. For breast ultrasound, AI identifies lesions missed by the human eye, aiding in early intervention. In UAE hospitals, this could cut diagnostic delays, vital as chronic diseases rise 15% annually per WHO data.

• Step 1: Image acquisition via diverse scanners.
• Step 2: Pretraining on OpenUS for feature learning.
• Step 3: Fine-tuning DB-SAM for segmentation.
• Step 4: Clinical integration with doctor oversight.

MBZUAI's Partnerships Driving UAE Healthcare AI

MBZUAI collaborates with Cleveland Clinic Abu Dhabi and Department of Health - Abu Dhabi (DOH) to deploy AI in clinics. Initiatives include AI Innovation Day for Health, showcasing tools for maternal care and robotics. These align with UAE's UAE Centennial 2071 vision for AI-led healthcare, investing AED 112 billion in digital health.

Qin's work integrates with FetalCLIP (210k fetal ultrasound images) and MediX-R1 (open-source RL for multimodal medical AI), amplifying impact. Faculty like Yaqub lead MICCAI 2026 in Abu Dhabi, positioning UAE as AI-medical hub.

Challenges in AI Medical Imaging and Solutions

Challenges include data scarcity, privacy (GDPR-compliant), and bias. OpenUS mitigates via public diverse data; DB-SAM's adapters reduce retraining needs by 90%. Ethical AI at MBZUAI ensures transparency, with human-in-loop for decisions.

Risks: Overreliance (5-10% error in edge cases), addressed by hybrid systems. UAE's regulatory sandbox tests safely.

Implications for UAE Higher Education and Research

MBZUAI's model—fully funded PhDs, industry ties—inspires UAE unis like Khalifa University. It produced 100% employed grads, many in healthcare AI. As UAE hosts 50% MENA AI talent, such programs boost GDP by AED 112bn via AI by 2031.

Explore UAE university jobs or higher ed positions for AI roles.

Photo by Vitaly Gariev on Unsplash

Future Outlook: AI's Next Frontier in UAE Medicine

Qin eyes postdoc blending imaging with EHRs for holistic diagnosis. MBZUAI plans multimodal datasets, 6G-AI integration. With UAE's AI healthcare market at $2bn by 2026, these innovations promise equitable, efficient care.

Stakeholders: Doctors gain tools; patients faster diagnoses; UAE global leadership.