AI Detects Brain Injury in Critically Ill Children: AWS and UPMC Develop Early Detection Solution
Measuring Brain Injury in Critically Ill Children: How AI Helps
AWS demonstrated the use of FHIR resources and AI to measure and detect brain injury in critically ill children.
Summary
University of Pittsburgh Medical Center (UPMC) partnered with Amazon Web Services (AWS) to develop an innovative solution that uses electronic health record data and machine learning for early detection of brain injury. The system aims to identify the risk of brain injury in critically ill pediatric patients before any physical symptoms appear.
Key Players
Details
This Proof of Concept (PoC) leverages various clinical information from Electronic Health Record (EHR) data, including medications and vital signs. The solution builds a pipeline that ingests structured data—specifically FHIR resource types such as patient, observation, encounter, and medication administration—into AWS HealthLake. Data is initially stored in CSV format on S3 buckets, then converted to FHIR R4 ndjson format via AWS Batch. This processed data is imported into HealthLake. The integrated data can be queried using Amazon Athena or used for training AI models with services like Amazon SageMaker AI. Crucially, the solution supports real-time inference by utilizing an API Gateway and Lambda function, allowing it to potentially hook directly into the EHR system. This technology aims to enable early and accurate intervention for critically ill children, representing a significant advancement in pediatric care.
Technology Note
FHIR（Fast Healthcare Interoperability Resources）は医療データ交換の国際標準。このエントリの関連技術: FHIR
Original content copyright by respective publishers