Integrating an AI platform into clinical IT: BPMN processes for clinical AI model development

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical IT Integration

Background:

• Growing use of Artificial Intelligence (AI) in hospitals has led to isolated solutions not integrated into clinical IT.
• A need exists for a clinical AI platform to manage the AI model development lifecycle within clinical IT infrastructure.

Purpose of the Study:

• To investigate the integration of a clinical AI platform into existing clinical IT infrastructure.
• To outline the stages of the AI model development cycle within clinical IT.
• To illustrate system landscape interactions using Business Process Model and Notation (BPMN) diagrams.

Main Methods:

• Conducted a requirements analysis for the clinical AI platform, considering clinical IT specifics.
• Identified and mapped processes for the entire AI model development cycle using BPMN diagrams.
• Evaluated identified processes with three clinical use cases using the FEDS framework.

Main Results:

• Developed BPMN process diagrams covering the full clinical AI model development cycle: Data Selection, Data Annotation, On-site Training and Testing, and Inference (Batch and Real-Time).
• Demonstrated through three clinical use cases that the approach accommodates a broad spectrum of healthcare AI applications.
• Successfully evaluated processes, confirming the comprehensive nature of the developed approach.

Conclusions:

• The clinical AI platform is suitable for local AI model development within clinical IT.
• The BPMN diagrams effectively cover diverse clinical AI use cases.
• This framework supports future advancements like multi-site AI model training and deployment, and enhanced security/privacy integration.