Researchers at Liwa University's Research Institute for AI and Emerging Technology in Abu Dhabi have made headlines with a groundbreaking publication in Nature Scientific Reports. The study introduces an advanced computer-aided diagnosis (CAD) system designed to detect and grade age-related macular degeneration (AMD), a major cause of vision loss worldwide. This two-stage AI model combines precise image segmentation with robust classification, achieving accuracies over 97 percent on a dataset of 864 retinal images. Led by Mohammed Ghazal, who holds positions at both Liwa University and Abu Dhabi University, the work underscores the UAE's growing prowess in AI-driven medical research.

This achievement not only highlights Liwa University's commitment to innovative research but also positions it as a key player in addressing healthcare challenges through higher education initiatives. As the UAE advances its Vision 2031 agenda, emphasizing AI and emerging technologies, such contributions from local institutions like Liwa are pivotal in bridging academia and practical health solutions.

Liwa University: A Hub for AI Innovation in the UAE

Established in 1993, Liwa University has evolved into a prominent higher education institution with campuses in Abu Dhabi and Al Ain. Offering accredited programs across the College of Engineering and Computing, College of Medical and Health Sciences, College of Media and Public Relations, and General Education, the university serves a diverse student body from over 30 nationalities. Its Research Institute for AI and Emerging Technology exemplifies Liwa's focus on cutting-edge fields, fostering collaborations that translate academic research into real-world applications.

In the context of UAE higher education, Liwa stands out for its industry-aligned curricula and 100 percent scholarships for Emirati students in health sciences through partnerships like Nafis. The institute's work on the AMD CAD system reflects broader national efforts to integrate AI into healthcare, supported by initiatives from the Ministry of Education and the UAE's National AI Strategy 2031.

Understanding Age-Related Macular Degeneration (AMD)

Age-related macular degeneration, often abbreviated as AMD, is a progressive eye condition that affects the macula, the central part of the retina responsible for sharp, detailed vision. It manifests in two forms: dry AMD, characterized by drusen deposits and gradual thinning of the macula, and wet AMD, involving abnormal blood vessel growth that leaks fluid and causes rapid damage. Globally, AMD impacts over 196 million people, projected to rise to 288 million by 2040 due to aging populations.

In the UAE, AMD prevalence stands at approximately 0.91 percent, with the therapeutics market valued at US$54 million in 2022 and expected to double by 2030. Factors like increasing life expectancy, diabetes rates, and UV exposure exacerbate risks in the region. Early detection is crucial, as timely interventions like anti-VEGF injections for wet AMD can preserve vision. Traditional diagnosis relies on optical coherence tomography (OCT) scans and fundus photography, but manual interpretation is time-consuming and prone to human error—challenges AI aims to overcome.

The Need for AI in AMD Diagnosis within UAE Healthcare

The UAE's healthcare system, renowned for its advanced infrastructure, faces rising demands from an aging expatriate and local population. AMD contributes significantly to blindness among those over 50, with wet AMD accounting for most severe cases. Current screening methods, while effective, struggle with scalability in busy clinics like those in Abu Dhabi. AI-powered tools promise faster, more accurate triage, freeing ophthalmologists for complex cases.

• Reduces diagnostic time from minutes to seconds per scan.
• Enhances consistency, minimizing inter-observer variability.
• Supports remote screening in underserved areas like Al Ain.
• Aligns with UAE's digital health strategy, including the AI in Health program.

Liwa's CAD system directly addresses these gaps, leveraging publicly available retinal datasets augmented with clinical data for robust performance.

Breakdown of the Two-Stage CAD System

The Liwa team's innovation lies in a hybrid two-stage pipeline. Stage one employs SegNet-MobileNet, a fusion of SegNet's encoder-decoder architecture for pixel-wise segmentation and MobileNet's lightweight convolutional backbone for efficiency. This delineates AMD lesions—drusen, fluid, and neovascularization—from healthy tissue in fundus images.

1. Preprocessing: Images resized to 224x224 pixels, normalized, and augmented (rotation, flipping) to handle variability.
2. Segmentation: SegNet-MobileNet outputs binary masks isolating regions of interest (ROIs).
3. Feature Extraction: From ROIs, 100+ features including texture (GLCM), shape, and intensity statistics are computed.
4. Classification: Ensemble ML models (CatBoost, XGBoost, Extra Trees) classify as normal, dry AMD, or wet AMD.

This modular design ensures both precision in lesion identification and generalizability across diverse retinal imaging conditions.

Photo by Zulfugar Karimov on Unsplash

Performance Metrics and Validation

Tested on 864 retinal images with demographic and clinical metadata, the system achieved remarkable results. Ensemble classifiers reached 97.8% accuracy, 98.2% precision, 97.5% recall, and 97.9% F1-score. Binary classification (AMD vs. normal) hit 99.1% accuracy. Comparative analysis outperformed baselines like ResNet50 (94.2%) and standard SegNet (92.7%).

Cross-validation confirmed robustness, with SHAP visualizations revealing key features like lesion area and texture entropy driving decisions.

Explainability: Building Trust with SHAP Analysis

Unlike black-box models, Liwa's system incorporates SHAP (SHapley Additive exPlanations) for interpretability. SHAP values quantify feature contributions, showing that ROI-based texture features dominate predictions. For instance, high entropy in segmented drusen regions strongly indicates dry AMD. This transparency is vital for clinical adoption, allowing doctors to validate AI outputs against patient history.

In UAE hospitals, where AI integration is accelerating, explainable AI fosters physician confidence and regulatory compliance under the UAE AI Ethics Principles.

Implications for UAE's Higher Education and Healthcare Landscape

This publication elevates Liwa University within UAE's competitive higher education sector, alongside giants like NYU Abu Dhabi and Khalifa University. It exemplifies how private institutions contribute to national priorities like the UAE Centennial 2071, focusing on health tech innovation. For students in Liwa's Engineering and Health Sciences programs, such research offers hands-on exposure to AI applications, boosting employability in Abu Dhabi's booming medtech sector.

• Potential integration into UAE's Sehhaty app for nationwide screening.
• Training data localization to capture diverse UAE demographics.
• Partnerships with Cleveland Clinic Abu Dhabi for clinical trials.

Collaborative Research: Bridging UAE and Global Expertise

The project's international flavor—USA and UAE authors—mirrors UAE's strategy of attracting global talent. Mohammed Ghazal's dual affiliations highlight inter-university synergies. Future expansions could involve UAEU or Masdar Institute, aligning with the Abu Dhabi AI Campus.

Stakeholders praise the work: UAE health officials note its alignment with the National Program for Chronic Diseases, while academics see it as a model for publishable, impactful research from emerging institutions.

Challenges and Future Directions

While promising, the study calls for multi-center validation to ensure generalizability across ethnicities prevalent in the UAE. Challenges include dataset diversity and real-time deployment on edge devices for clinics. Liwa plans enhancements like multimodal inputs (OCT + fundus) and federated learning for privacy-preserving training.

Looking ahead, this could evolve into a prognostic tool predicting AMD progression, integrating with wearables for proactive care.

Photo by Igor Rusakov on Unsplash

UAE's AI Research Ecosystem: Liwa's Strategic Role

The UAE invests heavily in AI, with $20 billion committed by 2031. Liwa's institute positions the university to secure grants from MBZUAI or ADIA Lab. For higher ed, it inspires curricula updates, like AI minors in health sciences, preparing graduates for roles at Sheikh Shakhbout Medical City or beyond.

Real-world cases: Similar AI tools at Moorfields Eye Hospital (UK) reduced diagnosis time by 30 percent; Liwa's could replicate this in UAE polyclinics.

Empowering the Next Generation of UAE Researchers

Liwa's student involvement in such projects—through capstone theses and internships—cultivates talent. Emirati scholarships ensure nationals lead, supporting UAE's knowledge economy. Actionable insights: Aspiring researchers should master PyTorch for segmentation and scikit-learn for ensembles, as used here.

As AMD cases rise with UAE's silver economy, Liwa's tool offers timely solutions, blending education, research, and societal impact.