FSH net

In today's healthcare, having a user-friendly central system to record clinic operations can greatly improve the quality of life in hospitals. Notably, FHIR (Fast Healthcare Interoperability Resources) is rapidly emerging as an integral component in hospitals.

Importing medical information(symptoms, surgery complications,…) from formats like PDFs or Excel files into a unified system like FHIR is a big challenge. Converting these documents is a time-consuming endeavour, characterized by the daunting task of extracting relevant information and mapping concepts or synonyms to their corresponding SNOMED CT identifiers.

Structuring free text narratives is cumbersome for humans but thats exactly where GPT outperforms

Enter GPT, an innovation that has gained widespread attention over the past year.  Its exceptional ability to comprehend context within documents makes it a useful tool to tackle this challenge. GPT can decipher contextual information embedded within these documents, seamlessly transforming them into FHIR formats such as ValueSets or CodeSystems. The missing piece of the puzzle lies in connecting these extracted concepts with their official SNOMED identifiers.

Fortunately, REST APIs offer a solution, enabling automated searches for concepts in FHIR terminology servers to retrieve their associated SNOMED IDs. By orchestrating a conversation between GPT and these REST APIs, a comprehensive solution emerges. This is precisely where FSH Net steps in.

FSH Net, a tool produced by our intern Tom Nissens, bridges the gap between GPT and our own FHIR terminology server. It facilitates the extraction of FSH formats from free text through this dynamic conversation. FSH (FHIR Shorthand) is notably more human-readable than raw FHIR syntax, making it an ideal intermediate format for GPT-generated content. This approach is valuable because while GPT is highly proficient, errors can occur. With FSH Net, a human can review and refine the output, ensuring its accuracy and coherence.

Check out FSH net: https://fshnet.streamlit.app/