Testing an artificial intelligence algorithm for detecting newborn hip dysplasia on ultrasound scans

Developing and testing computer-assisted diagnostic tools for screening of developmental dysplasia of the hip in newborns: a multi-reader multi-case study

Status

Recruiting

Phases

Unknown

Study type

Observational

Source

ISRCTN

Registry ID

ISRCTN49436239

Enrollment

Registered

2024-06-19

Start date

2024-03-01

Completion date

Unknown

Last updated

2024-07-08

For informational purposes only — not medical advice. Sourced from public registries and may not reflect the latest updates. Terms

Conditions

Developmental dysplasia of the hip in newborns, diagnosed by ultrasound scan Neonatal Diseases

Interventions

A retrospective dataset of 70 newborn ultrasound scans will be compiled to include 35 normal and 35 abnormal (dysplastic [25]/dislocated [10]) hips. The case balance is intended to better mimic clinic

Eligibility

Sex/Gender

All

Inclusion criteria

Inclusion criteria: Consultants/attendings (specialising in Paediatric Orthopaedic Surgery) and registrars/residents. Specialist physiotherapists who take part in hip screening as part of their clinical practice.

Exclusion criteria

Exclusion criteria: Any healthcare professional who does not review newborn hip ultrasound scans (either autonomously or under direct supervision) in their clinical practice

Design outcomes

Primary

Measure	Time frame
Reader and AI algorithm performance will be evaluated as sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and Area Under Receiver Operating Characteristic Curve (AUC). Where the hip is abnormal on the ultrasound and readers correctly identify this classification as abnormal, it will be counted as a true positive, an incorrect diagnosis of normal by the reader will be a false negative. Where the hip is normal on the ultrasound, its correct classification by the reader will be a true negative and an incorrect classification will be a false positive. The performance measures listed above will be compared for each reader with and without AI assistance. The performance of the AI algorithm alone will also be evaluated as a comparative measure.	—

Measure

Time frame

Reader and AI algorithm performance will be evaluated as sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and Area Under Receiver Operating Characteristic Curve (AUC). Where the hip is abnormal on the ultrasound and readers correctly identify this classification as abnormal, it will be counted as a true positive, an incorrect diagnosis of normal by the reader will be a false negative. Where the hip is normal on the ultrasound, its correct classification by the reader will be a true negative and an incorrect classification will be a false positive. The performance measures listed above will be compared for each reader with and without AI assistance. The performance of the AI algorithm alone will also be evaluated as a comparative measure.

—

Secondary

Measure	Time frame
Reader speed will be evaluated as the mean review time per scan, with and without AI assistance. Reader confidence will be evaluated via a self-reported score (scale of 1 to 5, 1= not confident to 5 = fully confident), with and without AI assistance.	—

Countries

England, United Kingdom

Contacts

Public ContactAbhinav Singh

Abhinav.singh@ndorms.ox.ac.uk+44 (0)1865227374

Outcome results

None listed

Source: ISRCTN (via WHO ICTRP) · Data processed: Feb 4, 2026