[Paper Review] Foundation Model for Advancing Healthcare: Challenges, Opportunities, and Future Directions

Yuting He, Fuxiang Huang+5 more|arXiv (Cornell University)|Apr 4, 2024

Primary Care and Health Outcomes5 citations

TL;DR

A comprehensive survey of healthcare foundation models (HFMs) across language, vision, bioinformatics, and multimodal sub-fields, detailing methods, data resources, applications, challenges, and future directions.

ABSTRACT

Foundation model, which is pre-trained on broad data and is able to adapt to a wide range of tasks, is advancing healthcare. It promotes the development of healthcare artificial intelligence (AI) models, breaking the contradiction between limited AI models and diverse healthcare practices. Much more widespread healthcare scenarios will benefit from the development of a healthcare foundation model (HFM), improving their advanced intelligent healthcare services. Despite the impending widespread deployment of HFMs, there is currently a lack of clear understanding about how they work in the healthcare field, their current challenges, and where they are headed in the future. To answer these questions, a comprehensive and deep survey of the challenges, opportunities, and future directions of HFMs is presented in this survey. It first conducted a comprehensive overview of the HFM including the methods, data, and applications for a quick grasp of the current progress. Then, it made an in-depth exploration of the challenges present in data, algorithms, and computing infrastructures for constructing and widespread application of foundation models in healthcare. This survey also identifies emerging and promising directions in this field for future development. We believe that this survey will enhance the community's comprehension of the current progress of HFM and serve as a valuable source of guidance for future development in this field. The latest HFM papers and related resources are maintained on our website: https://github.com/YutingHe-list/Awesome-Foundation-Models-for-Advancing-Healthcare.

Motivation & Objective

• Summarize the current progress of healthcare foundation models (HFMs) across four sub-fields: language, vision, bioinformatics, and multimodal.
• Analyze data, algorithmic, and computing infrastructure challenges hindering HFM deployment.
• Provide a taxonomy of pre-training and adaptation methods used in HFMs.
• Identify datasets, applications, and emerging directions to guide future research.

Proposed method

• Provide systematic taxonomy of HFMs by sub-field (LFM, VFM, BFM, MFM) and by pre-training paradigm (generative, contrastive, hybrid, supervised).
• Review adaptation strategies (fine-tuning, adapter tuning, prompt engineering) and their role in transferring HFMs to healthcare tasks.
• Summarize datasets and resources across four sub-fields to assess data availability and limitations.
• Survey applications and map latest HFMs to clinical contexts to illustrate practical impact.
• Discuss key challenges in data quality/diversity, algorithmic reliability, and computing infrastructure.

Experimental results

Research questions

• RQ1 What is the current progress of HFMs in healthcare across language, vision, bioinformatics, and multimodal domains?
• RQ2 What are the main data, algorithmic, and infrastructure challenges facing HFMs in healthcare?
• RQ3 What future directions and opportunities look most promising for advancing HFMs in clinical practice?

Key findings

• HFMs across four sub-fields have accelerated healthcare AI by enabling generalist capabilities applicable to diverse tasks.
• There are significant challenges related to data ethics, diversity, heterogeneity, and cost that impede large-scale HFM training and deployment.
• Computing infrastructure requirements for high-dimensional healthcare data are substantial and environmentally impactful.
• A broad taxonomy of pre-training and adaptation methods reveals widespread use of generative and contrastive learning in pre-training, and fine-tuning or prompt-based adaptation in downstream tasks.
• The survey provides extensive datasets, applications, and a forward-looking set of directions to guide future development in HFMs.