[Paper Review] Transformers in Healthcare: A Survey

Subhash Nerella, Sabyasachi Bandyopadhyay+10 more|arXiv (Cornell University)|Jun 30, 2023

Artificial Intelligence in Healthcare and Education12 citations

TL;DR

A survey of how Transformer architectures are applied to diverse healthcare data (imaging, EHR, social media, signals, and biomolecular sequences) and the associated benefits, limitations, and ethical considerations.

ABSTRACT

With Artificial Intelligence (AI) increasingly permeating various aspects of society, including healthcare, the adoption of the Transformers neural network architecture is rapidly changing many applications. Transformer is a type of deep learning architecture initially developed to solve general-purpose Natural Language Processing (NLP) tasks and has subsequently been adapted in many fields, including healthcare. In this survey paper, we provide an overview of how this architecture has been adopted to analyze various forms of data, including medical imaging, structured and unstructured Electronic Health Records (EHR), social media, physiological signals, and biomolecular sequences. Those models could help in clinical diagnosis, report generation, data reconstruction, and drug/protein synthesis. We identified relevant studies using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. We also discuss the benefits and limitations of using transformers in healthcare and examine issues such as computational cost, model interpretability, fairness, alignment with human values, ethical implications, and environmental impact.

Motivation & Objective

• Summarize how Transformers are used across healthcare data modalities (imaging, structured/unstructured EHR, social media, physiological signals, biomolecular sequences).
• Identify clinical applications enabled by Transformers such as diagnosis support, report generation, data reconstruction, and drug/protein synthesis.
• Discuss methodological considerations, including computational cost, interpretability, fairness, alignment with human values, and ethical/environmental impacts.

Proposed method

• Review literature using PRISMA guidelines to identify relevant transformer-based healthcare studies.
• Categorize applications by data modality and task (diagnosis, generation, reconstruction, synthesis).
• Discuss benefits, limitations, and practical considerations of transformer models in healthcare.
• Highlight issues of interpretability, fairness, alignment with human values, ethics, and environmental impact.

Experimental results

Research questions

• RQ1 What data modalities have transformers been applied to in healthcare?
• RQ2 What clinical tasks do transformers support in healthcare (e.g., diagnosis, report generation, data reconstruction, synthesis)?
• RQ3 What are the benefits, limitations, and ethical/environmental considerations of using transformers in healthcare?

Key findings

• Transformers have been applied to medical imaging, EHR data (structured and unstructured), social media, physiological signals, and biomolecular sequences.
• Applications include clinical diagnosis support, report generation, data reconstruction, and drug/protein synthesis.
• The survey discusses costs, interpretability, fairness, alignment with human values, ethical implications, and environmental impact in healthcare transformers.