A framework for generative AI-driven extraction of clinical user needs in pediatric device development.

Area of Science:

• Medical Device Development
• Artificial Intelligence in Healthcare
• Human-Computer Interaction

Background:

• Generative artificial intelligence (GenAI) and large language models (LLMs) streamline medical product development by automating tasks like data annotation and insight extraction from expert interviews.
• Traditional methods for analyzing interview data are time-consuming and labor-intensive, leading to administrative burdens and reduced efficiency in identifying clinical needs and device specifications.

Purpose of the Study:

• To explore the application of GenAI, specifically GPT-4o, in extracting user functional and design requirements from medical professional interviews.
• To facilitate the iterative development of an infant heart rate detector for neonatal resuscitation by leveraging AI-driven data analysis.
• To demonstrate the potential of LLMs in curating user requirements and design specifications for medical devices.

Main Methods:

• Conducted semistructured interviews with 29 healthcare practitioners, recording and transcribing a total of 26 hours of data.
• Utilized GPT-4o to extract user insights and requirements from interview transcripts.
• Compared GenAI-extracted insights with manual interviewer notes to assess accuracy and reliability.

Main Results:

• Validated the clinical need for a rapid and accurate heart rate measurement device during neonatal resuscitation among all interviewees.
• Extracted and curated user requirements categorized into ease of use, measurement accuracy and speed, reusability, display, battery life, start-up time, and cost.
• Performed quantitative analyses on interviewee experience, clinical settings, and specialties.

Conclusions:

• GenAI, particularly GPT-4o, shows significant potential in accurately and reliably extracting user requirements from medical interviews, reducing workload and improving productivity.
• The study provides a framework for researchers and product designers to utilize LLMs for curating user requirements and design specifications in medical device development.
• The findings support the iterative development of medical devices by efficiently identifying and organizing critical user-centered design criteria.