Workshop on
Deep Generative Models for Health

Deep generative models have recently gained unprecedented attention following recent advancements in text-to-image generation, diffusion models and large language models. Additionally, early well-established  approaches, such as variational autoencoders, generative adversarial networks, and normalizing flows, are widely applied for learning interpretable representations, as well as integrating multiple modalities or prior information from domain knowledge. These advancements hold the premise of unlocking significant potential in the health sector. 

Generative AI emerges as a compelling solution in addressing the challenges posed by the scarcity of medical datasets due to complex data acquisition processes and privacy regulations, the demand for accountable and interpretable methodologies, and the need to integrate multiple and diverse modalities. Despite the recently witnessed methodological advances, generative approaches are limited in their current real-world medical applications. This is arguably due to several open challenges that include designing objective validation procedures, as well as finding reliable metrics for learnt representations to assess interpretability and semantic content.

In this workshop, we provide a unique venue for the most recent trends in research on deep generative models, focusing on exploring their potential for health applications. We also provide the optimal setting to discuss the open problems that prevent these methods from having a profound positive impact in clinical settings. This workshop will be the ideal venue to attract a diverse pool of researchers aiming to integrate generative models in health scenarios.

We encourage submissions that leverage the recent methodological advancements in generative models to address critical medical challenges across all data-types, paving the way for their practical integration into the healthcare system. 

The workshop will include talks from world-renowned researchers with remarkable expertise on both methodological aspects of generative models, and their applications in the health domain. Our panel discussion will focus on the challenges and promises of generative AI in health. With the unifying goal of boosting the applicability of deep generative models for health data, researchers will share their lessons learnt across different application scenarios and methodological focus areas. This venue will serve as the basis to build a community to tackle the challenges preventing generative models from being widely applied in clinical settings, and elevate their impact on the future of healthcare. 

DGM4H Workshop - NeurIPS 2023