April 2026

OPHTHALMOLOGY

AI MODEL SUPPORTS MEDICAL LITERATURE REVIEW

Large language models (LLMs) are increasingly shaping how clinicians and scientists synthesize medical evidence. A recent Nature Communications study featuring Manjot Gill, MD, introduces a foundation model to support human-AI collaboration with potential implications for evidence-based medicine across specialties.

The Challenge: Scaling Evidence Synthesis
Systematic reviews and meta-analyses are foundational to clinical decision-making, yet they are time- and resource-intensive. As the volume of published clinical research continues to grow, traditional approaches to literature review can limit how quickly new evidence is translated into practice.

A New Approach: Human-AI Collaboration
The study team presents LEADS, an AI foundation model trained on large-scale, domain-specific medical data to support literature mining, including:
  • Study identification and screening
  • Data extraction from clinical trial publications

Unlike general-purpose LLMs, LEADS is designed for medical evidence synthesis and is evaluated within clinician- and researcher-led workflows.

Key Findings
Across multiple literature-mining tasks, the model demonstrated improvements over existing LLMs when used in collaboration with medical experts. In user studies involving clinicians and scientists, LEADS was associated with better recall and accuracy while reducing time spent on evidence review tasks.

Tools like LEADS may help clinicians stay current with research and support guideline development and clinical decision-making.

This work highlights how carefully trained, domain-specific AI models can complement physician expertise. 
Manjot Gill, MD, Vice Chair of Clinical Performance, Department of Ophthalmology, and Professor of Ophthalmology and Medical Education at Northwestern Medicine

Refer a Patient

Northwestern Medicine welcomes the opportunity to collaborate with you in caring for your patients. ​

You May Also Like