Country / Region
EMEA
Tags
Artificial intelligence, Data analytics, EHDS (European Health Data Space), Innovation, Research, Tooling
Imagine hospital notes transforming themselves into research-grade evidence before the night shift ends: built in Oxford, Striata, a SNOMED-powered AI Scientist ingests raw text, de-identifies it, and maps every concept to SNOMED CT (and companion vocabularies) in real time. Once structured, its multi-agent engine sweeps across diagnoses, medications, age, comorbidities, seasonal trends and workflow metrics, plotting hidden patterns, proposing testable hypotheses, running statistics and flagging bias, then hands users publication-ready reports. Early pilots, including Latvia's NHS, show Striata compresses weeks of analyst labour into minutes, turning any ward into a discovery hub without adding to workloads. By fusing large language model reasoning with SNOMED's formal logic it delivers faster scientific discovery, rapid pharmacovigilance, sharper equity audits and instant operational feedback.
Description
Oxford-based Amorphous AI has developed Striata, a cutting-edge multi-agent AI system designed to autonomously transform raw clinical text into actionable insights in real-time. Striata's cooperative AI agents seamlessly perform patient data de-identification, precise SNOMED CT mapping of clinical entities, and execute automated research cycles that include:
* Data Exploration ‚ Systematic and deep analysis of clinical data across demographics, timeframes, and disease states.
* Dynamic Visualization ‚ Real-time graphical representations that reveal subtle or complex clinical patterns.
* Hypothesis Generation ‚ Automatic identification of clinically significant associations, such as medication risks and operational inefficiencies.
* Statistical Testing & Bias Detection ‚ Robust validation processes, including successful autonomous detection of intentionally hidden synthetic drug-outcome associations in a specially constructed benchmark dataset.
* Automated Reporting ‚ Generation of detailed, regulator-compliant reports covering methodologies, findings, statistical rigor, and identified limitations.
In this project, we generated a synthetic dataset containing ground truths, such as associations between drugs and outcomes. We will present results of this benchmark, showcasing Striata's ability to mine complex health data for insights on par with world-class research teams at a fraction of the cost.
Scope
SNOMED's rich semantic structure provides ideal support for sophisticated, algorithmic reasoning.
Internationally recognized standard ensuring interoperability across global healthcare and regulatory bodies.
Essential compatibility with EHDS and FHIR standards, enabling healthcare providers to comply seamlessly with emerging European regulations.
Frequent and comprehensive updates allow for rapid integration of novel clinical insights and therapeutic interventions.
How SNOMED CT will be used
Canonical Encoding ‚ Clinical entities, including diagnoses, medications, procedures, and observations, are standardized to SNOMED identifiers, maintaining nuance through advanced post-coordination techniques.
Ontology-Based Reasoning ‚Leverages SNOMED's semantic hierarchy and causal relationships to dynamically refine patient cohorts without human-driven rules.
Quality Control with Snowstorm ‚Ensures accuracy and quality control of SNOMED mappings, automatically identifying and rectifying discrepancies following SNOMED updates.
Why SNOMED CT will be used
Contact
