BILD

Agenda

Summer School on Biomedical Imaging with Deep Learning (BILD)

Thursday 04.09

Wednesday 03.09

Tuesday 02.09

Friday 05.09

Monday 01.09

Welcome Note

Medical Image Segmentation Dr. Aditya Rastogi

Graph Intelligence Asso. Prof. Islem Rekik

09:00 - 10:30

Excursion

Keynote Speech Prof. Nassir Navab

10:30 - 11:00

Coffee break

Coffee break

Coffee break

An excursion to Sidi Bou Said and Carthage which was founded in the 9th century B.C. on the Gulf of Tunis, grew into a major Mediterranean trading empire and a center of a thriving civilization

Vision Language Models Dr. Aditya Rastogi

DL Fundamentals Prof. Shadi Albarqouni

Uncertainty Estimation Prof. Vasileios Belagiannis

11:00 - 12:30

Prof. Hanen Boussi

Prof. Ezzeddine Zagrouba

12:30 - 13:00

Uni. Tunis El Manar

Dr. Narjes Ben ameur

Visit the main Campus Networking Coktail

Lunch break Networking

Lunch break Networking

Lunch break Networking

13:00 - 14:30

Hands-on Session Medical AI Dev. with MONAI - NVIDIA

Hands-on Session Classification

Hands-on Session Segmentation

14:30 - 16:00

Coffee break

Coffee break

Coffee break

16:00 - 16:30

Hands-on Session Detection

Hands-on Session Quality Control

Hands-on Session Medical AI App Dev.

16:30 - 18:00

Welcome Note

• Prof. Moez Chafra

President (Rector) of the University of Tunis El Manar, Tunisia

• Prof. Monia Najjar

Vice-rector for International Cooperation and Strategic Directions at the University of Tunis El Manar, Tunisia

• Prof. Shadi Albarqouni

Professor of Computational Medical Imaging Research at Universitätsklinikum Bonn and University of Bonn, Germany

Keynote

Sensing, Perception, Intelligence & Action in High Intensity, Dynamic Surgical Environments

Over the past decade, the rapid advancements in machine learning have revolutionized various fields, significantly impacting our lives. In this talk, we will delve into the realm of medical applications and explore the challenges and opportunities associated with integrating these cutting-edge technologies into computer-assisted interventions. Our primary focus will be on fostering acceptance and trust in machine learning and robotic solutions within the medical domain, which often necessitates the path through Intelligence Amplification (IA). Augmented Reality allows us to leverage IA to augment human intelligence and expertise, ultimately paving the way for the seamless integration of Artificial Intelligence (AI) and robotics into clinical solutions. Drawing from some groundbreaking research conducted

at the Chair of Computer-Aided Medical Procedures at both TU Munich and Johns Hopkins University, I will present a series of novel techniques developed to address the unique demands of medical applications. Specifically, we will explore their practical implementations in diverse areas, including Robotic Ultrasound Imaging, Multimodal Data Analysis, and Semantic Scene Graphs for Holistic Modeling of the Surgical Domain. Furthermore, I will showcase compelling examples of how Augmented Reality solutions can serve as catalysts for embracing AI in computer-assisted surgery. In this talk, I will navigate at the intersection of machine learning, computer vision, advanced visualization, and medical robotics to travel along the path from intelligence amplification to artificial intelligence in computer-assisted interventions.

Hands-On Medical AI App Development: Pneumonia Detection from Chest X-ray

Dr. Lara Reimer

In this 1.5-hour hands-on workshop, participants will bring together the results of the previous sessions to create a fully functional, cross-platform mobile application for pneumonia detection. Designed for students in computer science and medicine, the workshop bridges the gap between algorithm development and real-world application. Participants will implement an end-to-end workflow: capturing or uploading images of chest X-ray, applying image pre-processing techniques, and running the data through a machine learning model to detect signs of pneumonia. The final app will present the classification results in a user-friendly interface, enabling anyone—regardless of technical background—to test the model in practice. By the end of the session, students will have collaboratively built a prototype that demonstrates how medical AI can be translated into usable digital health tools.

Medical Image Seg.

This lecture introduces students to the principles and methodologies of medical image segmentation, a critical task in computer-assisted diagnosis and treatment planning. We begin by motivating the role of segmentation in clinical applications, followed by an overview of classical and modern segmentation algorithms, including thresholding, clustering, atlas-based methods, and deep learning-based approaches. The lecture explores standard evaluation metrics such as Dice and Jaccard coefficients, and provides insight into the limitations posed by scarce, weak, or noisy annotations. Finally, we highlight state-of-the-art strategies to overcome these challenges through data augmentation, transfer learning, semi-supervised methods, and post-processing techniques.

Learning Outcomes:

• Understand the motivation and clinical relevance of semantic segmentation in medical imaging.
• Compare classical and modern segmentation algorithms
• Describe and compute evaluation metrics
• Recognize key challenges in medical image segmentation

• Apply and evaluate modern solutions

Vision Language Models

Transformers have revolutionized natural language processing and are now reshaping the landscape of computer vision and multimodal learning. This talk explores the evolution of transformer-based models from early attention mechanisms to powerful vision-language architectures like CLIP and MedCLIP. Emphasizing practical applications over theoretical deep dives, we discuss how vision transformers (ViT) are adapted for image recognition, how natural language supervision enables zero-shot learning, and how domain-specific adaptations like MedCLIP overcome the limitations of generic models in medical imaging. The presentation also touches on key challenges in transfer learning, dataset construction, prompt engineering, and fine-tuning in high-stakes domains like radiology

Learning Outcomes:

• Understand the motivation and explain the core architecture of transformers and how they differ from CNNs in vision tasks.

• Describe the architecture and training objectives of ViT and CLIP, and their advantages and limitations compared to CNN-based models.
• Discuss how contrastive learning enables zero-shot transfer.
• 5. Analyze the role of prompt engineering and data scale in achieving robust generalization across tasks and domains.

BILD Summer School

BILD 2025 is organized under the umbrella of the Strategic Arab-German Network for Affordable and Democratized AI in Healthcare (SANAD), uniting academic excellence and technological innovation across borders. The Albarqouni Lab organizes this year’s edition at the University Hospital Bonn and the University of Bonn. We are proud to partner with leading institutions in the region—Lebanese American University, University of Tunis El Manar, and Duhok Polytechnic University — to deliver a truly international learning experience.

Graph Intelligence:

From Learning to Reasoning

This lecture offers a gentle introduction to Geometric Deep Learning, focusing on Graph Neural Networks (GNNs) and Graph Convolutional Networks (GCNs). We will cover learning on both homogeneous and heterogeneous graphs, with applications in brain connectivity, medical imaging, and beyond. The session will highlight how graph-based models can capture complex relationships in data and discuss key challenges such as scalability and explainability

Learning Outcomes:

• Understand the core principles of GNNs
• Learn how GNNs operate on homogeneous and heterogeneous graphs
• Explore applications in brain connectivity, medical imaging, and healthcare, and
• Identify key challenges and active research directions in the field

Sidi Bou Said

Perched on a cliff overlooking the Gulf of Tunis, Sidi Bou Said enchants visitors with its iconic blue-and-white façades, winding cobblestone lanes, and postcard-perfect panoramas This fairytale-like village, named after the revered Sufi scholar Abu Said al-Baji, has become a magnet for artists and intellectuals—its charming fusion of Arabic-Andalusian architecture and creative spirit echoes through every corner. A sensory feast awaits at hidden tea houses and sunset cafés like Café des Délices—here, jasmine-scented breezes, Mediterranean views, and local artistry come together in an unforgettable day-trip getaway