Design implications of repurposing a radiomics research platform for education: The case of QuantImage

Abstract

Purpose This paper addresses the complex relationships between AI, medical education and research, embodied in QuantImage v2, a software platform for radiomics. Existing research has indicated lack of user studies and physicians’ low confidence in the diagnosis and treatment planning proposed by AI models. In this context, we explore how QuantImage, designed as a tool for radiomics research, can be repurposed for education. Methods Specifying the possible educational uses and required technical adjustments of QuantImage, we have organized user studies in the form of collective trial sessions at the Centre hospitalier universitaire vaudois (CHUV) in Lausanne, Switzerland. These 13 sessions, in which pairs of novice users worked with the platform together with an expert tutor, have been video recorded, transcribed, and analyzed in detail, focusing on troubles that participants encountered. Results Based on the analyses, we formulate actual and potential design implications. Actual changes already implemented in the platform include a paradigm shift in feature selection and highlighting central elements of the user interface, which are both motivated by the aims of making the work with QuantImage more accessible to users with varying levels of expertise. Potential improvements include implementation of three different modes for the use of the platform: basic, advanced, and tutorial mode, with the last one reflecting specific needs of higher education and broadening the relevance of the obtained skills beyond the particular platform. Conclusions Understanding the practical work with AI-based models is key for making radiomics respectable in oncology and radiology and overcoming the resistance to their adoption in clinical practice. Taking into account that different user groups and use scenarios place different requirements on the design of QuantImage opens the possibility of using the platform to leverage timely reflections of the practical role of AI in medicine.

Ähnliche Arbeiten