Artificial Intelligence-Assisted Matching of Human Postmortem Donors to Ocular Research Projects

Area of Science:

• Ophthalmology
• Biomedical Informatics
• Artificial Intelligence

Background:

• Scarcity of human ocular samples with short postmortem intervals (PMIs) hinders ophthalmic research and drug discovery.
• Manual matching of donor data by eye banks is time-consuming, inefficient, and error-prone.
• Unstructured donor medical data in free text fields limits interoperability with matching databases.

Purpose of the Study:

• To test the ability of artificial intelligence (AI) to structure unstructured donor medical data for improved matching.
• To enhance the ReSync semi-automated matching system by incorporating AI (ReSyncAI).

Main Methods:

• A retrospective study utilizing a large language model with natural language processing to structure donor medical data.
• Secure transfer of historical donor medical data to the AI model for structuring and standardization.
• Integration of structured data back into the ReSync system for analysis and match testing.

Main Results:

• Achieved a 94.2% success rate in medical terminology keyword extraction, correction, and standardization.
• Successfully structured donor data demonstrated full interoperability with the ReSync system.
• ReSyncAI accurately matched donors to "age-related macular degeneration" despite abbreviations, misspellings, and incomplete data.

Conclusions:

• AI effectively structures unstructured donor medical data, enabling seamless integration with matching systems like ReSync.
• ReSyncAI significantly improves the efficiency and accuracy of matching ocular donors for research.
• This AI-driven approach addresses a critical bottleneck in ophthalmic research and drug discovery by optimizing sample utilization.