An evaluation of a checklist in musculoskeletal radiographic image interpretation when using artificial intelligence

Abstract

INTRODUCTION: Artificial intelligence (AI) is being used increasingly in image interpretation tasks. Human reliance on technology and bias can cause decision errors. A checklist, used with the AI to mitigate against such biases, may optimise the use of AI technologies and promote good decision hygiene. A checklist to aid radiographic image interpretation for radiographers using AI for image interpretation was formed. This study investigates the effect of a checklist for musculoskeletal (MSK) radiographic image assessment when using AI interpretive assistance. METHODS: Radiographers were asked to interpret five MSK examinations with AI feedback. They were then provided with the checklist and asked to reinterpret the same five examinations with the AI feedback (n = 140 interpretations). During the interpretation sessions, participants were asked to provide a diagnosis and a confidence level on the diagnosis provided. Participants were then asked to complete a questionnaire to gain feedback on the use of the checklist. RESULTS: Fourteen radiographers were recruited. Nine participants found the checklist alongside the AI most useful and five participants found the AI element to be most useful on its own. Five participants found the AI feedback to be useful as it helped to critique the radiographic image interpretation more closely and rethink their own initial diagnosis. CONCLUSION: The checklist for use with AI in MSK image interpretation contained useful elements to the user, but further developments can be made to enhance its use in clinical practice.

Ähnliche Arbeiten