Skip to content

jeremiah-louis/Solacea

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

23 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

RAGHack: Colon Cancer Decision Support System using Microsoft GraphRag

Solacea Medium Banner

Objective

Our MVP aims to provide doctors with actionable insights to aid in adjuvant chemotherapy (ACT) decision-making for patients with stage II/III colorectal cancer. Using ctDNA (circulating tumor DNA) data from clinical studies, our system helps guide treatment recommendations, focusing on precision, early intervention, and personalized care.

Problem Statement

Colorectal cancer is a leading cause of cancer-related deaths worldwide. For patients in stage II/III colorectal cancer, the decision to administer adjuvant chemotherapy is crucial but often complex. Current diagnostic methods primarily rely on imaging and physical assessment, which may not fully capture the risk of recurrence.

Circulating tumor DNA (ctDNA) has emerged as a highly sensitive biomarker that can detect molecular residual disease (MRD) post-surgery, signaling the likelihood of cancer recurrence. However, interpreting ctDNA results and integrating them into treatment decisions requires advanced tools to assist doctors.

Challenge

Doctors need a data-driven decision support system to:

  • Interpret ctDNA results.
  • Assess the benefit of adjuvant chemotherapy.
  • Make informed, personalized treatment decisions based on clinical evidence.

Solution: The RAG-Based System

Our MVP leverages Microsoft’s GraphRAG approach to deliver targeted, contextual insights by querying medical research and clinical guidelines. Initially, the system uses data from the BESPOKE CRC study, which provides essential findings on the role of ctDNA in guiding chemotherapy decisions.

Image Description

How It Works

  1. Input: Doctors input relevant patient information (e.g., stage of cancer, ctDNA status).
  2. Data Retrieval: The system queries specific research papers, clinical trials, and medical guidelines related to colorectal cancer treatment, focusing on ctDNA's predictive value.
  3. Augmentation: Using a combination of retrieved research and contextual understanding, the system generates a personalized treatment recommendation.
  4. Output: The doctor receives a tailored recommendation on whether to pursue adjuvant chemotherapy based on the patient’s ctDNA status and the best available clinical evidence.

Visualised Graph

Screenshot 2024-09-18 at 11 32 28 The graph includes all the communities, text units and entities derived from the data and was visualised using https://noworneverev.github.io/graphrag-visualizer/

Example Query

Input:
"A Stage III patient is ctDNA positive post-surgery. What does the research suggest regarding adjuvant chemotherapy?"

Output:
"For Stage III ctDNA-positive patients, adjuvant chemotherapy significantly improves disease-free survival (DFS) based on the BESPOKE CRC study. Chemotherapy is recommended to reduce recurrence risk."

Key Features

  • Focused Data Scope: The MVP uses BESPOKE CRC study data, ensuring precision and depth in ctDNA-related decision-making.
  • Microsoft GraphRAG: Employs Microsoft’s advanced retrieval-augmented generation approach to derive insights from clinical data and research.
  • Targeted Recommendations: Provides personalized guidance based on the patient’s stage, ctDNA status, and existing clinical studies.
  • Expandable: Future iterations will incorporate more research, case studies, and clinical guidelines for broader applicability.

Technology Stack

  • Microsoft GraphRAG: For advanced retrieval and augmentation.
  • Python: For backend logic and integration.
  • Streamlit: For creating a simple, interactive frontend interface.
  • Lancedb: For vector-based data retrieval.
  • Docker: For containerizing the application, ensuring portability.

Data Source

  • Primary Data: Circulating tumor DNA (ctDNA) for informing adjuvant chemotherapy (ACT) in stage II/III colorectal cancer (CRC): Interim analysis of BESPOKE CRC study.
  • Future Data: Other relevant colorectal cancer clinical trials and studies focusing on ctDNA and chemotherapy outcomes.

Use Case Scenarios

Scenario 1: Patient Monitoring

A doctor inputs the ctDNA status of a post-surgery Stage II patient. The system recommends whether adjuvant chemotherapy is necessary, based on the patient's risk of recurrence and clinical data from the BESPOKE CRC study.

Scenario 2: Treatment Adjustment

For a ctDNA-positive patient undergoing chemotherapy, the doctor queries whether continuing treatment is beneficial. The system provides insights based on data regarding ctDNA clearance and recurrence patterns from clinical studies.

Next Steps

  • Data Expansion: Incorporate additional clinical studies and guidelines to improve the range of recommendations.
  • User Feedback: Gather feedback from doctors during the MVP phase to improve the interface and recommendation accuracy.
  • Clinical Validation: Collaborate with medical experts to ensure the system’s recommendations align with clinical best practices.

Conclusion

Our MVP is a lightweight, focused tool designed to support doctors in making informed decisions about adjuvant chemotherapy in colorectal cancer patients, based on ctDNA insights. By starting small with a specific data set, we aim to create a scalable solution that can expand to incorporate a wider body of clinical evidence over time.

About

A decision support system leveraging Microsoft GraphRAG to assist doctors in determining the need for adjuvant chemotherapy for stage II/III colorectal cancer patients. The system uses ctDNA data and clinical research to provide personalized, data-driven treatment recommendations.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors