BreastCare Trials 🌸

🌟 Inspiration

Breast cancer is one of the most prevalent forms of cancer worldwide, accounting for 11.7% of all new cancer cases (Global Cancer Observatory, 2022). Despite advances in medical science, heterogeneity in molecular subtypes, varying stages of diagnosis, and diverse therapy responses pose significant challenges for 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 often struggle to access and interpret critical clinical trial information.
• ⏳ Time Sensitivity: Delays in finding relevant clinical trial data can hinder timely treatment decisions.
• 👩‍⚕️ Personalization: Clinical trial data is often generalized and not tailored to specific patient profiles.

These barriers hinder innovation and limit the impact of clinical trial data, making it difficult to maximize its potential in 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 for breast cancer and its end users—clinicians, researchers, and patients. By combining cutting-edge AI technologies, it provides precise, actionable insights to revolutionize breast cancer care.

✨ Key Features:

• 🔎 Streamlined Search: Leverages Snowflake Cortex Search for precise and efficient retrieval of extensive National Cancer Institute (NCI) clinical trial data.
• 📝 AI-Driven Summaries: Generates concise, human-readable insights tailored to user queries using Mistral-large2.
• 👩‍⚕️ Patient-Centric Insights: Tailors clinical trial recommendations and summaries based on specific patient profiles and queries.
• 💻 Interactive Frontend: A user-friendly interface built with Streamlit, ensuring intuitive navigation for medical professionals and patients alike.
• 📈 Real-Time Updates: Provides up-to-date clinical trial information, ensuring users access the latest data.

🔧 How I built it

🛠 Tech Stack:

• ❄️ Snowflake Cortex Search: Enables fast, precise retrieval of clinical trial data.
• 🧠 Mistral-large2: Generates context-aware summaries to simplify complex information.
• 🌐 Streamlit: Delivers an interactive and accessible frontend.
• 📑 NCI Clinical Trial Data: Powers the platform with clinical trial documents for breast cancer.

🛠 Development Process:

• 🔗 Model Integration: Seamlessly combined Snowflake Cortex Search with Mistral-large2 for robust RAG capabilities.
• 🎨 User-Centric Design: Focused on building an intuitive interface that caters to both medical professionals and patients.

⚡ Challenges we ran into

• 🤔 Integrating Snowflake Cortex Search with Mistral-large2 for efficient RAG implementation.
• 🎛️ Designing a user-friendly interface that balances accessibility and professional utility.
• 🕒 Ensuring real-time updates and handling large clinical trial datasets.

🏆 Accomplishments that I am proud of

• 🎉 Successfully combining advanced AI models for context-aware information retrieval and summarization.
• 🙌 Building an intuitive, patient-centric platform for empowering users with actionable insights.
• ⏱️ Achieving real-time performance with low latency and high relevance in clinical trial data retrieval.

📚 What I learned

• The importance of optimizing chunk size and overlap to balance latency, cost, and relevance.
• The challenges of processing and summarizing complex medical data for varied user needs.
• Insights into user-centric design principles for healthcare applications.

🚀 What's next for BreastCare Trials

• 🌐 Expanding Data Sources: Integrating additional clinical trial datasets for broader coverage.
• 📊 Enhanced Personalization: Incorporating advanced profiling for more tailored recommendations.
• ⚙️ Performance Optimization: Further reducing latency and improving context relevance.
• 🤝 Collaborative Features: Adding tools for collaboration among healthcare professionals and patients.

Built With

• mistral-large2
• snowflake
• snowflake-cortex-search
• streamlit