User Manual

1. Introduction

The Clinical EMR Extraction Workbench is designed to assist clinicians and researchers in rapidly structuring unstructured medical data (e.g., free-text clinical notes, pathology reports). It leverages advanced Large Language Models (LLMs) to read textual data and populate structured registries.

2. Clinical Records Extraction

This module handles standard patient encounters and clinical notes.

• Extract Record: Paste or upload a patient's raw clinical note. The system will process it against configured LLM prompts to extract demographics, history, staging, and outcome metrics.
• Review Cases: After extraction, the record appears in the "Cases" list. Click on a case to review the structured output.
• Verification: Ensure you verify the structured fields. The LLM acts as an assistant, but clinical validation is required before registry export.

3. Pathology Reports

Dedicated tools for batch processing and structured extraction of pathology data.

• Upload Reports: Upload single or multiple pathology PDF/Text files.
• Review Reports: Check the extracted biomarkers, margin status, and histological grades.
• Export Data: Select validated reports and export them as a CSV file for your institutional database.

4. Administration (Admins Only)

System administrators can configure prompts and user access.

• Prompt Configuration: Modify the system prompts sent to the LLM to tune the extraction guidelines. (E.g., instructing the LLM on specific TNM versions).
• User Management: Create, edit, or disable user accounts. Manage Roles (Admin vs User).
• Settings: Configure LLM backend endpoints (e.g., OpenAI, Azure, Local Ollama) and parameters like temperature and max tokens.

5. Security & Privacy

Important: Ensure that no Protected Health Information (PHI) is sent to external APIs unless your institution has a BAA in place with the provider, or you are using a secure, local, on-premise model.