🎓 Bouthaina Slika defends her PhD on AI for Pulmonary Disease Assessment!

🎉 Congratulations to Bouthaina Slika on the successful defense of her PhD thesis,
“Advancing Pulmonary Disease Assessment in Medical Imaging: Deep Regression and Data Augmentation Techniques”, on Tuesday, July 22, 2025 (10:30 am) at the University of the Basque Country (UPV/EHU)!

Her thesis was supervised by Dr. Fadi Dornaika and Dr. Karim Hammoudi.

🧠 Thesis Contributions

Bouthaina’s PhD work focuses on developing deep learning models and novel data augmentation methods for accurate and efficient assessment of pulmonary diseases using chest X-rays (CXR) and CT scans. Her research addresses critical challenges in COVID-19, pneumonia, and related pulmonary infections, making strides toward practical, clinically relevant AI systems.

Key contributions include:

• Multi-Score Prediction of Lung Infection Severity
  A novel multi-task deep learning algorithm combining a dual transformer encoder and feature fusion module. This method simultaneously predicts infection spread and opacity scores, improving accuracy and outperforming state-of-the-art baselines (IEEE TETCI, 2025).

• Parallel VMamba + Attention Model
  A cutting-edge approach using the VisualStateSpace (VSS) framework and segmented lung replacement augmentation, enabling robust pneumonia severity prediction from CXRs with excellent generalization (Diagnostics, 2025).

• PViTGAtt-IP Architecture
  A parallel vision transformer model with cross-attention for severity quantification in both CXRs and CT scans, validated across multiple datasets (IEEE Transactions on Big Data, 2025).
  🔗 Code on GitHub

• Lightweight Vision Transformer for CXR Severity
  Developed ViTReg-IP, a transformer-based regressor requiring fewer parameters but maintaining high accuracy and robustness across diverse datasets (MBEC, 2024).
  🔗 Code on GitHub

• Multi-task Vision Transformer (MViTReg-IP)
  Introduced a method for predicting infection extent and opacity simultaneously, improving diagnostic reliability (Conference paper, 2024).
  🔗 Code on GitHub

• SuperpixelGridMasks Augmentation
  Co-developed a new data augmentation technique based on superpixel decomposition, now publicly available on GitHub, showing strong improvements across biomedical and real-world datasets (Journal of Healthcare Informatics Research, 2022).
  🔗 Code on GitHub

Her work demonstrates how deep regression models and innovative augmentations can deliver reliable severity assessments, providing doctors with fast, objective tools for clinical decision-making.

🧑‍⚖️ Defense Committee

The committee for her defense was composed of:

• Dr. Jinan Charafeddine (Chair, Pôle Universitaire Léonard de Vinci, France)
• Dr. Blanca Rosa Cases Gutiérrez (Secretary, UPV/EHU, Spain)
• Dr. Denis Hamad (Voting member, Université du Littoral Côte d’Opale, France)

👏 Bravo, Bouthaina! Your research paves the way for AI-powered healthcare tools that can enhance diagnosis, improve patient outcomes, and prepare medicine for future pandemics. We wish you the best for your next steps! 🚀