Overview

Schema Version

https://atlas.rsna.org/schemas/2025-11/model.json

Name

Impact of Different Artificial Intelligence User Interfaces on Lung Nodule and Mass Detection on Chest Radiographs

Link

https://dx.doi.org/10.1148/ryai.220079

Indexing

Keywords: Artificial Intelligence, Chest Radiograph, Conventional Radiography, Lung Nodule, Mass Detection

Content: CH

RadLex: RID10345, RID39056, RID50149

Author(s)

Jennifer S. N. Tang

Jeffrey K. C. Lai

John Bui

Wayland Wang

Paul Simkin

Dayu Gai

Jenny Chan

Diane M. Pascoe

Stefan B. Heinze

Frank Gaillard

Elaine Lui

Organization(s)

Department of Radiology, Royal Melbourne Hospital

Department of Radiology, University of Melbourne

Version

1.0

License

Funding

Authors declared no funding for this work.

Ethical review

Approved by The Royal Melbourne Hospital Ethics Committee; case data were retrospectively collected and de-identified; patient consent waived; written informed consent obtained from all reader participants.

Date

Updated: 2023-05-01

Published: 2023-03-22

Created: 2022-04-25

References

[1] Tang JSN; Lai JKC; Bui J; Wang W; Simkin P; Gai D; Chan J; Pascoe DM; Heinze SB; Gaillard F; Lui E. "Impact of Different Artificial Intelligence User Interfaces on Lung Nodule and Mass Detection on Chest Radiographs". Radiology: Artificial Intelligence. 2023;5(3):e220079. 2023-03-22. doi:10.1148/ryai.220079. PMID: 37293345. PMCID: PMC10245182.

Model

Architecture

Commercial comprehensive deep-learning model for chest radiographs (Annalise.ai) producing 124 clinical findings; study varied user interface components (text, AI confidence score, image overlay).

Availability

Commercially available via Annalise.ai platform; demo interface customized for the study. https://annalise.ai/

Clinical benefit

When presented as text-only output, AI assistance significantly improved radiologist detection of lung nodules/masses on chest radiographs compared with no AI.

Clinical workflow phase

Clinical decision support during image interpretation.

Degree of automation

Assists radiologists with decision support (text output, optional confidence score and image overlay) without automating final diagnosis.

Indications for use

Assisting radiologists in detecting lung nodules and masses on chest radiographs in a hospital radiology setting.

Input

Chest radiograph images presented via a web-based interface.

Limitations

Single-center retrospective paired-reader study with small participant and case numbers; UI options limited to those available on a single proprietary software; potential selection bias; fixed limited viewing time per case (25 seconds) may have constrained use of more complex UIs; presence of other pathologies may have affected performance.

Output

CDEs: RDE1704.1, RDE1702.3

Description: Three UI variants of AI output: (a) text-only; (b) text + AI confidence score; (c) text + AI confidence score + image overlay (localization). Confidence score and overlay displayed only when AI predicted a positive finding.

Recommendation

UI selection for AI outputs should be validated against performance, not solely user preference.

Regulatory information

Comment: Per article, Annalise.ai chest radiograph algorithm has TGA and CE clearance.

Authorization status: Cleared (Therapeutic Goods Administration [Australia] and European CE marking) for the underlying Annalise.ai chest radiograph algorithm.

Use

Intended: Detection

User

Intended: Radiologist