BreastCare Trials 🌸

🌟 Inspiration

Breast cancer remains one of the most prevalent cancers worldwide, accounting for 11.7% of all new cancer cases (Global Cancer Observatory, 2022). Despite medical advancements, challenges such as molecular subtype heterogeneity, varying diagnosis stages, and diverse therapy responses create barriers to achieving uniform patient outcomes.

🚩 Challenges in Breast Cancer Care:

• 📊 Data Overload: Clinical trials generate vast, complex datasets that are difficult to navigate and utilize effectively.
• 🔍 Accessibility Gap: Clinicians, researchers, and patients struggle to access and interpret critical clinical trial information.
• ⏳ Time Sensitivity: Delays in finding relevant clinical trial data hinder timely treatment decisions.
• 👩‍⚕️ Personalization: Clinical trial data is often generalized, lacking specificity for individual patient profiles.

These barriers limit innovation and prevent the full potential of clinical trial data from advancing breast cancer care.

🤖 What It Does

BreastCare Trials is an AI-powered Retrieval-Augmented Generation (RAG) application designed to bridge the gap between complex clinical trial data and its end users—clinicians, researchers, and patients. Using cutting-edge AI technologies, it delivers precise, actionable insights to revolutionize breast cancer care.

✨ Key Features:

• 🔎 Streamlined Search: Utilizes Snowflake Cortex Search for fast and accurate retrieval of National Cancer Institute (NCI) clinical trial data.
• 📝 AI-Driven Summaries: Generates concise, human-readable insights using Mistral-large2 tailored to user queries.
• 👩‍⚕️ Patient-Centric Insights: Provides personalized clinical trial recommendations and summaries.
• 🖥 Interactive Frontend: Built with Streamlit to offer an intuitive and accessible user experience.
• 📈 Real-Time Updates: Ensures users have access to the latest clinical trial information.

🛠️ How It Was Built

🛠️ Tech Stack:

• ❄️ Snowflake Cortex Search: Enables precise and efficient retrieval of clinical trial data.
• 🧠 Mistral-large2: Generates context-aware summaries for simplified interpretation.
• 🌐 Streamlit: Provides an interactive and user-friendly frontend.
• 📝 NCI Clinical Trial Data: Serves as the primary data source for breast cancer clinical trials.

🎯 Development Process:

• 🔗 Model Integration: Seamlessly combined Snowflake Cortex Search with Mistral-large2 to enhance RAG capabilities.
• 🎨 User-Centric Design: Built an intuitive interface catering to both medical professionals and patients.

⚡ Challenges Faced

• 🤔 Integrating Snowflake Cortex Search with Mistral-large2 for efficient RAG implementation.
• 🎮 Designing a user-friendly yet professional interface.
• ⏳ Ensuring real-time updates while handling large datasets.

🏆 Accomplishments

• 🎉 Successfully integrated advanced AI models for intelligent data retrieval and summarization.
• 🙌 Built an intuitive, patient-centric platform for better decision-making.
• ⏱️ Achieved low-latency, high-relevance clinical trial data retrieval.

📚 Lessons Learned

• Optimizing chunk size and overlap improves latency, cost, and relevance.
• Processing complex medical data requires balancing technical accuracy with user accessibility.
• User-centric design is crucial for healthcare applications.

🚀 What's Next for BreastCare Trials

• 🌐 Expanding Data Sources: Integrating additional clinical trial datasets.
• 📊 Enhanced Personalization: Implementing advanced patient profiling.
• ⚙️ Performance Optimization: Further reducing latency and improving accuracy.
• 🤝 Collaborative Features: Enabling seamless collaboration among healthcare professionals and patients.

🛠️ Installation & Usage

1. Clone the repository: sh git clone https://github.com/your-repo/BreastCare-Trials.git
2. Navigate to the project directory: sh cd BreastCare-Trials
3. Install dependencies: sh pip install -r requirements.txt
4. Run the Streamlit app: sh streamlit run app.py
5. Access the application in your browser and explore the features!

Built With

• mistral-large2
• snowflake
• snowflake-cortex-search
• stremlit