International standards are essential for harmonising best practices to ensure the quality of products and services. IEC 62304 is a recognised reference standard for defining development life cycle (DLC) activities for medical device software (MDS). However, given the growing adoption of AI within this sector, it is essential to embed AI-specific requirements to guide the development of AI-enabled Medical Device Software (AIeD). As part of previous work, this paper reviews and maps IEC 62304:2006+A1:2015 and ISO/IEC 5338:2023, a process standard that provides generic life cycle processes for AI systems. These two standards are derived from ISO/IEC/IEEE 12207 and share elements that can be integrated to propose an initial DLC framework for AIeD. The integration in volves joining process IDs and identifying relevant keywords from each activity/task. Fourteen processes from ISO 5338 are aligned across activities in the IEC 62304 development process, including two AI-specific processes: Data Engineering and Knowledge Acquisition. Additionally, +11 AI-specific tasks and +24 AI-specific guidelines are identified and aligned. Future work will incorporate additional relevant standards and regulatory considerations to further detail and support this framework. Lessons from this paper could stimulate future research to automate the integration of standards, thereby enabling easier compliance for MDS manufacturers.
Zapata et al. (Fri,) studied this question.