Wholistic report generation for Breast ultrasound using LangChain

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Clinical Informatics

Background:

• Breast ultrasound (BUS) is crucial for breast abnormality detection.
• BUS report generation is time-consuming due to data integration needs.
• Current methods lack efficiency in synthesizing diverse clinical information.

Purpose of the Study:

• To explore the feasibility of a modular, AI-assisted framework for BUS report generation.
• To demonstrate the integration of AI tools within a clinical workflow.
• To improve the efficiency and consistency of BUS reporting.

Main Methods:

• Developed classification networks for BUS image analysis.
• Utilized LangChain to coordinate AI models with Large Language Models (LLMs).
• Integrated a Retrieval-Augmented Generation (RAG) component for patient data incorporation.

Main Results:

• The AI framework successfully generated structured, clinically meaningful BUS reports.
• The system demonstrated practical integration of image analysis and LLMs.
• Reports were consistent, interpretable, and aligned with radiologists' assessments.

Conclusions:

• The proposed modular AI framework offers a feasible solution for semi-automated BUS report generation.
• This approach provides a foundation for refining and potentially deploying AI in clinical reporting.
• The system enhances efficiency and consistency in breast imaging diagnostics.