Breast Cancer Detection in Microscope Images

One in eight women in the US will develop breast cancer, making early detection crucial. We've developed an LLM-based model to identify benign and malignant breast tumors in microscope images.

Comment 1

Please visit https://hatch2025-breakhis.web.app/ for demo

Inspiration We developed this project to reduce human error in breast cancer detection by creating an AI-powered companion app for microscopy image analysis.

What It Does One in eight women in the US will develop breast cancer, making early detection crucial. We've developed an LLM-based model to identify benign and malignant breast tumors in microscope images.

Development Process Data: Trained on the Breast Cancer Histopathological Database (BreakHis) Model: Utilized ChatGPT 4o for image analysis UI: Built using Google IDX for accessible implementation

Challenges Time constraints limited our original two-model approach:

  • An image classifier with 89% accuracy on BreakHis data
  • An LLM-based diagnostic model We initially planned to combine both models to reduce hallucination but couldn't fully integrate them within the hackathon timeframe.

Accomplishments & Learnings We gained deeper insights into breast cancer complexity and developed technology with significant real-world potential for improving diagnostic accuracy.

Future Development

  • Expand our dataset for increased accuracy
  • Complete the model integration to reduce AI hallucinations
  • Implement the solution in clinical settings

Built With

Share this project:

Updates