Impact of different artificial intelligence user interfaces on lung nodule and mass detection on chest radiographs
Author
Tang, Jennifer | The Royal Melbourne Hospital, Australia
Scientific poster presentation (W5A-SPCH-1) at RSNA 2023, 26 – 30. November 2023 in Chicago, US
Purpose
How do different AI user interfaces impact user preference and user performance in lung nodule and mass detection on CXR?
Method
140 cases ground-truthed for the presence of lung nodules or masses (81 negative, 59 positive) were reported by 10 participants using either no AI tool or an Annalise Enterprise CXR with one of three different user interface formats.
Results
AI text output significantly improved AUC for lung nodule detection, while AUCs did not differ for other output formats. 80% of users preferred the interface with text, overlay, and confidence bar.
Key Takeaway
AI output format may impact clinical performance while not necessarily matching with user preference.
Disclaimer
Harrison.ai Radiology Solutions were previously marketed as Annalise.ai solutions.
