This research presents a novel approach to classifying rare medical events in images where traditional deep learning fails due to limited data availability.

• Leverages Large Language Models to generate customized prompts for open-vocabulary image classification
• Creates domain-specific prompts that improve rare event detection without requiring large training datasets
• Demonstrates an effective zero-shot learning approach that enhances medical image classification accuracy
• Offers a scalable solution for medical conditions where collecting substantial training data is challenging

This innovation addresses a critical gap in medical diagnostics, enabling earlier detection of rare conditions when traditional ML approaches are impractical due to data scarcity.

Generating customized prompts for Zero-Shot Rare Event Medical Image Classification using LLM