AI solution tackles complex healthcare data for faster, more accurate diagnoses

Business Problem

With the primary business of providing end to end Healthcare Services, over 20+ year periods & plans, the client gathered a host of patient data, including medical histories, diagnostic images, lab results, and physician notes. The complexity and volume of this data made it difficult for healthcare professionals to quickly assimilate and interpret the information needed for accurate, real-time health assessments, which was critical to provide differentiated services to their patients. Traditional methods were slow and often failed to identify crucial correlations, leading to less informed clinical decisions.

About Atria

As an independent Healthcare Services provider, Atria is a budding leader in healthcare Services, with an extensive network of expert doctors and medical officers serving thousands of patients globally.

Visit Website

Explore Now

Solution

To address Atria’s challenges, GoML’s GenAI experts built a comprehensive solution that integrated advanced technologies, including knowledge graphs and GenAI models, to manage and analyze complex healthcare data, with multi-modal capabilities of Vision, Audio & text analysis.

The solution details are highlighted below:

Knowledge Graph Construction

Leveraged NebulaGraph and Weaviate to create a dynamic and scalable knowledge graph, enhancing semantic search capabilities and modeling complex relationships between health parameters, treatment histories, and patient outcomes.

Retrieval-Augmented Generation (RAG) Pipeline

Developed a RAG pipeline on AWS Bedrock to dynamically retrieve and update the knowledge graph, ensuring generative models like Claude V3, BioLLAMA2, and MedPalm2 had access to the most relevant and recent data for accurate health assessments.

Data Integration and Preprocessing

Utilized AWS Lambda for efficient, serverless data processing, ensuring seamless integration and real-time processing of data from various sources into Snowflake.

API Integration and Secure Delivery

Used FastAPI over AWS for robust API management, integrating with Google Docs and a dedicated Stream Channel for secure and efficient data transmission, ensuring seamless communication and access to health summaries by medical professionals and patients.

Generative AI Model Deployment

Deployed cutting-edge AI models, including Claude V3 for initial generation and BioLLAMA2 and MedPalm2 for specialized medical validations, to navigate the RAG-enabled knowledge graph and generate detailed, context-aware health summaries aligned with current medical standards.

Architecture

User Authentication: Okta manages user authentication and secure access control, integrating with the login module for user access.
Data Processing: AWS Lambda processes patient data from Snowflake, including notes, lab results, and scans, ensuring efficient data ingestion and handling.
Knowledge Graph and Search: NebulaGraph constructs a knowledge graph from patient data, while Weaviate enables efficient vector-based search and information retrieval.
Machine Learning Models: Utilizes OpenAI GPT-4 for generating insights, Google PaLM 2 (MedPalm 2) for medical validation, and Meta LLAMA 2 (OpenBioLLM) for biological data validation, ensuring accurate and relevant outputs.
Backend Services: Atria Server, powered by FastAPI and Docker, handles backend processing, business logic, and integrates with other services for scalable and consistent deployment.
API Management and Logic: API Gateway manages API requests, routing them to the backend, while logic layers encapsulate business rules and data validation processes.
Integration and Output: Processed data is integrated with Google Docs for document generation and shared via Stream Channel on the Atria Platform, providing users with accessible and actionable insights.