Artificial Intelligence in Endoscopy Research

Artificial Intelligence: Our View

As artificial intelligence rapidly diffuses into clinical medicine and research, we see substantial opportunity to leverage these tools to improve efficiency and reduce the administrative and ancillary burdens that often slow meaningful scientific work. The AI tools that can advance the efficiency and quality of clinical research are rapidly advancing and we aim to be at the forefront of these technologies. We also recognize that GI endoscopy continues to be a field where AI-based technologies can and should be rigorously studied as they are rapidly implemented into clinical practice.

At the same time, not all uses of AI align with our values. WIRE approaches AI as we approach all research tools: with curiosity, discipline, and clear boundaries. Below is a summary of how we think about AI in our lab.

Where AI Is Appropriate

Research Workflow Efficiency

WIRE aims to be on the forefront of leveraging AI in ways that reduce friction without replacing intellectual contribution. This includes tasks such as:

  • Constructing manuscript outlines to help organize ideas before drafting
  • Utilizing tested and validated technologies to assist in comprehensive literature searches
  • Summarizing peer reviewer comments to aid in revision planning
  • Collating author information and formatting requirements for journal submission
  • Populating checklists for CONSORT, STROBE, TARGET, or PRISMA compliance
  • Helping draft IRB templates or administrative boilerplate that do not involve scientific interpretation
  • Assisting in rewording of original language to meet word limit requirements (without changing the message)

Coding and Technical Development

With human oversight, AI-assisted coding may be used to hasten project development and conduct. At WIRE, we want to be leaders in the utilization of existing technology and the development of novel tools.

  • Developing internal tools that improve research efficiency
  • Exploring automated AI models that can translate clinical procedure reports to importable eCRF data for randomized trials
  • Building (HIPAA-compliant and IRB-approved) de-identification applications for procedure reports prior to eCRF import
  • Automating structured data extraction from standardized fields
  • Developing internal dashboards for project tracking
  • Checking and revising R coding for statistical analyses that are performed with human content expertise
  • Assisting in the design, build, and maintenance of wire-lab.org and other public-facing material

Idea Exploration and Concept Testing

AI may be used as an exploratory tool within various aspects of research, including:

  • Pressure-testing study designs
  • Identifying potential methodological blind spots
  • Generating alternative endpoint frameworks for internal discussion
  • Simulating and anticipating potential reviewer critiques

Where AI Is Not Appropriate

While the potential for AI to positively impact clinical research is immense, there are equally concerning aspects of AI which have already infected widespread clinical research. These are clear lines we do not cross. WIRE investigators will NEVER:

  • Use AI to generate original, unreviewed, and unrevised text for research manuscripts
  • Fail to disclose circumstances in a project or publication in which AI was leveraged
  • Generate sections of original scientific interpretation
  • Fabricate references, citations, or data
  • Allow AI tools to make unreviewed analytic or statistical decisions
  • Use AI to generate a list delineating our values on where AI is appropriate

Our Position

Scientific integrity is paramount. At WIRE, AI will be deployed to reduce administrative drag, improve workflow efficiency, and accelerate responsible research. It will not be used to outsource thinking, interpretation, or authorship.