Abstract
Abstract
Introduction: 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.
Methods: Radiographers were asked to interpret five Musculoskeletal (MSK) examinations. They were then provided with the checklist and asked to reinterpret the same five examinations with the AI feedback (n=140). 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: 14 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 AI feedback to be helpful 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.
Keywords: Artificial intelligence, checklist, image interpretation, musculoskeletal.
Introduction: 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.
Methods: Radiographers were asked to interpret five Musculoskeletal (MSK) examinations. They were then provided with the checklist and asked to reinterpret the same five examinations with the AI feedback (n=140). 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: 14 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 AI feedback to be helpful 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.
Keywords: Artificial intelligence, checklist, image interpretation, musculoskeletal.
| Original language | English |
|---|---|
| Publication status | Published (in print/issue) - 11 Oct 2024 |
| Event | The European Society of Medical Imaging Informatics - Pisa, Italy Duration: 11 Oct 2024 → 13 Oct 2024 |
Conference
| Conference | The European Society of Medical Imaging Informatics |
|---|---|
| Country/Territory | Italy |
| City | Pisa |
| Period | 11/10/24 → 13/10/24 |
