**************************************************************************************************
IJCNN 2025: Special Session on Foundation Models in Medicine (FMM)

Monday, June 30th to Saturday, July 5th, 2025, Rome, Italy

𝗢𝗿𝗴𝗮𝗻𝗶𝘇𝗲𝗿𝘀: Matteo Tortora, Aurora Rofena, Valerio Guarrasi

𝗦𝘂𝗯𝗺𝗶𝘀𝘀𝗶𝗼𝗻 𝗱𝗲𝗮𝗱𝗹𝗶𝗻𝗲: January 15, 2025
**************************************************************************************************

The special session on 𝗙𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗠𝗼𝗱𝗲𝗹𝘀 𝗶𝗻 𝗠𝗲𝗱𝗶𝗰𝗶𝗻𝗲 (𝗙𝗠𝗠) is part of the 𝗜𝗻𝘁𝗲𝗿𝗻𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗝𝗼𝗶𝗻𝘁 𝗖𝗼𝗻𝗳𝗲𝗿𝗲𝗻𝗰𝗲 𝗼𝗻 𝗡𝗲𝘂𝗿𝗮𝗹 𝗡𝗲𝘁𝘄𝗼𝗿𝗸𝘀 (𝗜𝗝𝗖𝗡𝗡 𝟮𝟬𝟮𝟱), which will be held at the 𝗣𝗼𝗻𝘁𝗶𝗳𝗶𝗰𝗮𝗹 𝗚𝗿𝗲𝗴𝗼𝗿𝗶𝗮𝗻 𝗨𝗻𝗶𝘃𝗲𝗿𝘀𝗶𝘁𝘆, 𝗥𝗼𝗺𝗲, 𝗜𝘁𝗮𝗹𝘆, from 𝗠𝗼𝗻𝗱𝗮𝘆, 𝗝𝘂𝗻𝗲 𝟯𝟬𝘁𝗵 𝘁𝗼 𝗦𝗮𝘁𝘂𝗿𝗱𝗮𝘆, 𝗝𝘂𝗹𝘆 𝟱𝘁𝗵, 𝟮𝟬𝟮𝟱.

𝗙𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗠𝗼𝗱𝗲𝗹𝘀 (𝗙𝗠𝘀) mark a breakthrough in deep learning, distinguished by their large-scale and task-agnostic design. Trained through self-supervised learning on vast and diverse datasets, these models acquire a broad understanding that can be applied to a range of downstream tasks.

In the medical field, FMs hold transformative potential, from clinical natural language processing and advanced computer vision for medical imaging analysis to multimodal data integration supporting personalized medicine, such as patient-specific Digital Twins. However, realizing their full impact requires addressing current limitations, particularly the technical and ethical challenges unique to medical data.

This session aims to create a collaborative forum to discuss recent advancements, share practical insights, and address the challenges researchers and practitioners face in applying FMs to medical applications. We invite submissions presenting novel methods, experimental findings, and theoretical contributions from both academia and industry. Accepted papers will undergo peer review and will be presented at the conference.

Topics include but are not limited to:
• 𝗣𝗿𝗲-𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝗶𝗲𝘀 𝗳𝗼𝗿 𝗠𝗲𝗱𝗶𝗰𝗮𝗹 𝗙𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗠𝗼𝗱𝗲𝗹𝘀: Techniques for training models on large-scale biomedical datasets, self-supervision, and adapting pre-trained models from general domains to specific medical applications.
• 𝗙𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗠𝗼𝗱𝗲𝗹𝘀 𝗳𝗼𝗿 𝗠𝗲𝗱𝗶𝗰𝗮𝗹 𝗜𝗺𝗮𝗴𝗶𝗻𝗴: Applications of FMs in radiology, pathology, and other medical imaging domains.
• 𝗙𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗠𝗼𝗱𝗲𝗹𝘀 𝗳𝗼𝗿 𝗠𝗲𝗱𝗶𝗰𝗮𝗹 𝗧𝗶𝗺𝗲 𝗦𝗲𝗿𝗶𝗲𝘀: Techniques for leveraging FMs to analyze time-series data such as physiological signals, vital signs, and ECG data for continuous monitoring, early diagnosis, and personalized treatment planning.
• 𝗟𝗮𝗿𝗴𝗲 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗠𝗼𝗱𝗲𝗹𝘀 𝗶𝗻 𝗛𝗲𝗮𝗹𝘁𝗵𝗰𝗮𝗿𝗲: Applications of LLMs for clinical decision support, patient engagement, and knowledge retrieval from electronic health records.
• 𝗠𝘂𝗹𝘁𝗶𝗺𝗼𝗱𝗮𝗹 𝗙𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗠𝗼𝗱𝗲𝗹𝘀 𝗶𝗻 𝗛𝗲𝗮𝗹𝘁𝗵𝗰𝗮𝗿𝗲: Approaches to integrate data from diverse sources (e.g., imaging, genomic, and clinical data) into FMs to achieve more holistic and accurate medical insights.
• 𝗜𝗻𝘁𝗲𝗿𝗽𝗿𝗲𝘁𝗮𝗯𝗹𝗲 𝗔𝗜 𝗶𝗻 𝗠𝗲𝗱𝗶𝗰𝗮𝗹 𝗙𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗠𝗼𝗱𝗲𝗹𝘀: Techniques for explaining complex model decisions and ensuring trust in clinical settings.
• 𝗘𝘁𝗵𝗶𝗰𝗮𝗹 𝗜𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀 𝗼𝗳 𝗙𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗠𝗼𝗱𝗲𝗹𝘀 𝗶𝗻 𝗛𝗲𝗮𝗹𝘁𝗵𝗰𝗮𝗿𝗲: Discussing ethical concerns such as data privacy and bias must be addressed to ensure fair and equitable healthcare for diverse populations.