Learning from Small Data in Medical Imaging (LSD-MI)

Overview

The integration of advanced learning techniques in medical imaging offers unique opportunities for improved diagnostic accuracy and personalized treatment. However, limited data availability poses significant challenges, particularly for models requiring large datasets, leading to reduced generalizability and increased overfitting risk. This special session will address these challenges and explore strategies for extracting insights from small datasets. We will discuss methodologies such as data augmentation, transfer learning, active learning, and expert-driven annotations to enhance model performance. By emphasizing robust validation and interpretability, this session aims to advance medical imaging despite data limitations.

The "Learning from Small Data in Medical Imaging (LSD-MI)" session aims to address the challenges of limited data availability in the medical imaging field. We invite discussions on innovative methodologies that enhance model performance, including data augmentation, transfer learning, expert-driven annotations, and active learning. Topics of interest encompass strategies for improving diagnostic accuracy, ensuring model generalizability, and fostering interpretability in data-scarce environments. By bridging the gap between advanced learning techniques and clinical relevance, this session seeks to engage the broader data mining community and provide valuable insights for ADMA2025 attendees focused on practical applications in healthcare.

Call for papers

We invite contributions that address aspects including, but not limited to:

• Techniques for data augmentation in medical imaging
• Transfer learning approaches for small datasets
• Few-shot learning methodologies tailored for medical applications
• Active learning strategies to optimize data collection
• Expert-driven annotation methods and their impact on model performance
• Uncertainty quantification methods in medical imaging
• Evaluating model generalizability in limited data scenarios
• Robust validation techniques in medical imaging
• Interpretability of machine learning models in healthcare
• Applications of large language models (LLMs) in medical imaging
• Case studies demonstrating successful applications of small data methodologies
• Ethical considerations and biases in small dataset research
• Collaborative frameworks for sharing medical imaging data across institutions

Submission information

• All submissions must follow the Submission Guidelines
• Number of pages are limited to 15 pages (LNAI formats)
• Papers are reviewed using "double blind" rule. Authors need prepare manuscript to fulifill these requirements.
• Submit your manuscript from here (CMT).

Important dates

• Submission deadline: 8 May 2025 5 June 2025 (extended)
• Review result notification: 27 July 2025
• Camera-ready deadline: 10 August 2025
• Conference scedule: 22-24 October 2025

Committee

Organizing Committee

• Essam A. Rashed, University of Hyogo, Japan – (Homepage)
• Alaa Tharwat, Bielefeld University of Applied Sciences and Arts (HSBI), Germany – (Homepage)

Technical Program Committee

• Tarek Gaber, University of Salford, UK
• Muhammad Nouman, University of Hyogo, Japan
• Ahmed Elboardy, University of Hyogo, Japan
• Mahmoud Bekhit, Australian Catholic University, Australia
• Ghada Khoriba, Nile University, Egypt
• Tamer Arafa, Nile University, Egypt
• Syed Ahmad Chan Bukhari, St. John’s University, USA
• Dina Elmanakhly, Suez Canal University, Egypt
• Haytham Ali, University of Tsukuba, Japan
• Mona Selim, Suez University, Egypt
• Syed Shamsul Islam, Edith Cowan University, Australia

Contact

For further details, please contact the organizing committee at lsdmi2025@gmail.com