Call for papers
Scope of the workshop
The Personalized Incremental Learning in Medicine (PILM) workshop at ACM Multimedia 2025 aims to explore the intersection of incremental learning and multimedia applications in the healthcare domain. As the paradigm of continual and incremental learning continues to advance, its integration with multimedia data presents unique opportunities and challenges. Indeed, in real-world contexts where data is rich, diverse, and continuously generated, models must adapt to new information while retaining previously learned knowledge. This is particularly crucial in personalized medicine, where machine learning models need to be tailored to individual patient characteristics, while facing challenges due to high-dimensional, sparse, and heterogeneous data sets, coupled with privacy concerns and the complexities of aggregating data from different sources. This workshop seeks to facilitate discussions on how recent advancements in continual and incremental can be applied to personalized medicine, leveraging multimedia data such as medical imaging, patient records, and biometric sensor data.
In particular, the PILM workshop will focus on bridging the gap between theoretical frameworks of incremental learning and their practical applications in personalized medicine using multimedia data, with the following objectives:
- To highlight the latest advancements in incremental learning that can enhance the development of personalized medical models using multimedia data.
- To foster discussions on unique challenges when applying incremental learning to patient-specific multimedia data with limited samples.
- To encourage the development of novel methodologies for integrating new multimedia patient data into existing models without compromising previous performance.
- To explore strategies for overcoming domain shifts within multimedia data, ensuring robustness across diverse medical equipment and imaging modalities.
- To promote the sharing of insights and techniques addressing data privacy challenges and the inaccessibility of historical patient multimedia data in continual learning contexts.
- To stimulate collaboration between researchers in machine learning, multimedia, and clinical practice to develop scalable, patient-centric solutions.
Topics
Submissions to the workshop are encouraged to address, but are not limited to, the following topics of interest:
- Novel algorithms for incremental and continual learning applicable to multimedia data in medical settings.
- Methods to prevent catastrophic forgetting in patient-specific machine learning models utilizing multimedia data.
- Strategies for one-shot or few-shot learning from multimedia data in medical diagnosis and treatment personalization.
- Techniques for handling domain shifts within multimedia data over time or across different medical devices.
- Approaches for integrating incremental learning with transfer learning in multimedia healthcare applications.
- Evaluation metrics and methodologies for assessing incremental learning systems’ performance in personalized medicine.
- Ethical considerations and data privacy solutions in developing incremental learning models using multimedia data.
- Case studies and practical applications of incremental learning in medical imaging, patient monitoring, and other areas of personalized medicine.
- Discussions on current datasets’ limitations and proposals for new multimedia data collection efforts supporting incremental learning research in medicine.
- Methods for fine-tuning foundation models to specific institutions or multimedia applications.
- Interdisciplinary research combining insights from clinical practice, multimedia, and machine learning to advance personalized medicine.