Building healthcare-context aware applications with HealthChain

Generative AI has the potential to improve care and increase productivity within the NHS, especially when integrated in existing workflows. However, getting started on projects like these can be complex and time-consuming, requiring expertise in both AI and healthcare systems.

In this talk I will introduce HealthChain, an open-source Python framework that makes it easy to prototype and test AI/NLP applications that integrate with electronic health record (EHR) systems by providing sandbox environments for end-to-end pipeline and integration testing without having to deal with the intricacies of backend APIs and data configurations. I will also discuss how this will help us build reactive, reproducible, and scalable systems in health beyond initial proof-of-concepts.

Transcript

1. HealthChain Jennifer Jiang-Kells Building healthcare- context aware applications with
2. Preprint About Me - Research software engineer specialising in NLP - Experience designing and implementing real-time AI solutions in NHS clinical settings - Spent too much time working with electronic health record systems (EHRs) Education 2021 - 2025 2024 - Open source developer tools for the next generation of healthcare AI applications
4. We want people to use our VeryUseful™ AI solution 👩💻 🍣
5. Healthcare organisations have standards and configurations 🏥 🍱
6. We need to make sure our solutions fit into these standards 🍣 🍱 ➡
7. We imagine the process to look like this 👩💻🍣 🏥🍱 🏥🍱 🏥🍱 🏥🍱 🏥🍱 🏥🍱 🏥🍱 ➡
8. But organisations are run by people 👩💻🍣 🧑⚕🍱 🧙🍱 🥷🍱 ➡
9. And people adapt standards 🧑⚕ 🍱 🧙🍱 🥷 🍱 👩🔬 🍱 👨🍳 🍱 🦹 🍱
11. So in reality it’s a bit more like this 👩💻🍣 ⏳ 🧑⚕ 🍱 🧙🍱 🥷 🍱 ⏳ ⏳
12. Things also get more complicated with generative AI 🤖 🎲 ✨ ❓ 🍘 🍙 🍣
13. Eventually we find ourselves juggling manual tasks 🤹 🍘 🍙 🍣 🍱 🍱 🍱
14. Can we make it be tt er?
15. An open-source Python framework to simplify building, testing and evaluating AI and NLP applications in a healthcare context HealthChain 💫 🏥
16. 🤖 Build pipelines integrated with NLP and ML libraries 🔌 Built-in CDA and FHIR connectors 🚀 Deploy servers locally with FastAPI Build An open-source Python framework to simplify building, testing and evaluating AI and NLP applications in a healthcare context HealthChain 💫 🏥
17. 🤖 Build pipelines integrated with NLP and ML libraries 🔌 Built-in CDA and FHIR connectors 🚀 Deploy servers locally with FastAPI Build 🧪 Test end-to-end pipelines in sandbox environment 🎨 Generate synthetic data for development 🔮 Version and track experiments Test An open-source Python framework to simplify building, testing and evaluating AI and NLP applications in a healthcare context HealthChain 💫 🏥
18. Building a Clinical Decision Support (CDS) application with HealthChain 💫 🏥 Summarising discharge encounters to streamline operational efficiency Example Use Case
19. Building a Clinical Decision Support (CDS) application with HealthChain 💫 🏥 Summarising discharge encounters to streamline operational efficiency Example Use Case Model Standardised API Con fi gs Data Versions
20. Building a Clinical Decision Support (CDS) application with HealthChain 💫 🏥 Summarising discharge encounters to streamline operational efficiency Example Use Case Model Standardised API Cards EHR Con fi gs Data Versions
21. Pipeline
22. Pipeline
23. doc.hl7.fhir_data doc.text JSON Request JSON Response
24. Pipeline
25. CdsFhirConnector Input Pipeline
26. CdsFhirConnector HFTransformer Input Summarization Pipeline
27. CdsCardCreator CdsFhirConnector HFTransformer Input Summarization Forma tt ing Pipeline
28. CdsCardCreator CdsFhirConnector HFTransformer Input Summarization Forma tt ing CdsFhirConnector Output Pipeline
29. dotimplement.github.io/HealthChain/cookbook/cds_sandbox/ Sandbox
30. dotimplement.github.io/HealthChain/cookbook/cds_sandbox/ @hc.api @hc.sandbox Sandbox
31. dotimplement.github.io/HealthChain/cookbook/cds_sandbox/ Sandbox @hc.api @hc.sandbox DataGenerator @hc.ehr
32. dotimplement.github.io/HealthChain/cookbook/cds_sandbox/ Sandbox @hc.api @hc.sandbox @hc.ehr DataGenerator
33. dotimplement.github.io/HealthChain/cookbook/cds_sandbox/ h tt p://0.0.0.0:8000/docs
34. An open-source Python framework to simplify building, testing and evaluating AI and NLP applications in a healthcare context Get involved! 💫 🏥 Democratise AI in Healthcare with open-source Reduce barriers to implementing AI in clinical settings Make evaluation of patient impact easier Lots more to do! 🚀
35. Thank you! Assets: Freepik Icons: Icons8 Linkedin Github ⭐ Star HealthChain on Github! 🤝 Connect with me on LinkedIn 🍱 🍣