CLIN07 - Comparing AI and manual organ measurements in abdominal ultrasound

Overview

Ultrasound is unusual among imaging modalities: the measurement and the image are made at the same moment, by the same person, in real time. Newer systems can now place some of those measurements automatically using on-board AI. This raises a practical question for clinical practice — how closely do the AI's measurements agree with those a qualified sonographer makes by hand, and where do they diverge?

This project extends work begun in last year's VRES, which contributed to a manuscript currently under review. Using a high-end commercial ultrasound platform, experienced sonographers will scan healthy volunteers and record organ measurements both manually and with the AI. The student will help quantify the agreement between the two approaches across a range of abdominal organs.

The topic is interdisciplinary, sitting across sonography, medical imaging, and the evaluation of clinical AI. Ethics approval will be in place prior to commencing.

Research engagement

Working alongside the supervisor and a small group of qualified sonographers, the student will be involved across the research process rather than in a single task. The student will not perform scanning; data are collected by the sonographers on consenting volunteers.

Research activities

Activities will include:

• A short scoping review of measurement-agreement methods and AI measurement in ultrasound
• Helping schedule and run data-collection sessions, and observing/assisting the sonographers during acquisition
• Recording and organising paired AI and manual measurements
• Cleaning the dataset and running basic agreement analyses (e.g. Bland–Altman, intraclass correlation) with support
• Contributing to a short write-up of the findings, including co-authorship

Research skills

Skills gained: reliability and agreement statistics, structured data handling, an introduction to analysis in R, an understanding of how clinical AI is evaluated rather than assumed, and practical experience of the research and scientific-writing process.

Outcomes

To quantify the level of agreement between AI-generated and manually acquired organ measurements in abdominal ultrasound, and to identify where the two approaches diverge. The work is expected to produce a clean dataset and analysis suitable for publication, and to lay the groundwork for a planned follow-up study examining how sonographers modify AI output in practice.

Skills and experience

Suited to a student enrolled in medical imaging, sonography, or a related health or science discipline, with an interest in AI in clinical practice. Some comfort with numbers is helpful but not essential — support is provided for the analysis. The ideal candidate is organised, reliable, and curious about how a clinical tool is tested rather than taken on trust. Prior scanning experience is not required, as the student does not perform ultrasound.

Start date

2 November, 2026

End date

19 February, 2027

Location

QUIQ  Lab  (Kelvin Grove Campus)

Additional information

Resources available to the student: structured supervision from the team; access to the QUIQ lab and its ultrasound equipment; the data‑capture workflow; secure QUT data storage; R and library/EndNote; an expert sonographer panel for the reference ratings

Keywords

• Ultrasound
• artificial intelligence
• measurement agreement
• medical imaging

Contact

Chris Edwards

3138 1924

c8.edwards@qut.edu.au