G230(P) Automated bone age assessment yields comparable results to paediatric radiologist reports. (12th March 2018)

• Record Type: Journal Article
• Authors: Jarvis, P; Kumar, Y; Giles, K; Thorogood, S
• Is Part Of: Archives of disease in childhood. Volume 103(2018)Supplement 1
• Page Start: A95
• Page End: A95
• Publication Date: 2018-03-12
• Subjects: Children -- Diseases -- Periodicals; Infants -- Diseases -- Periodicals; 618.920005
• DOI: 10.1136/archdischild-2018-rcpch.225
• Languages: English
• ISSNs: 0003-9888

Abstract

Abstract : Aims: To compare automated bone age assessment software with consultant paediatric radiologist reports in order to determine acceptability for use in our centre. Background: Bone age assessment is widely used in the evaluation of endocrine and genetic diseases as well as to assess the response to medical therapy. Our centre uses the Tanner and Whitehouse 2 radius, ulnar and short bones (TW2 RUS) method which involves grading these bones of the hand and wrist individually using maturity indicators from a reference book. This is then used to calculate a skeletal maturity score, which in turn is transformed into a bone age. This can be time consuming, especially for the less experienced reporter. It has also been shown that there is significant intra and inter-observer variability in assessment. Automated methods have been developed to reduce reporting time as well as observer variation. Methods: Every bone age x-ray (hand and wrist) examination over a 3 month period was both reported by a consultant paediatric radiologist and assessed by an automated method (BoneXpert; Visiana, Denmark). We then analysed the TW2 results of each to see if the automated method would be acceptable for use in our centre. Results The mean variability between reporter and automated method was 0.71 years with a standard deviation of 0.58 years. This compares favourably with published studies which have separately reported TW2 specific interobserver variability of 0.74 years and an interobserver standard deviation of 0.71 years. Our results also equate to a 9.18% mean variability (SD 7.84) between reporter and automated software. Conclusion: Automated bone age assessment software yields comparable results to consultant paediatric radiologist reports. In particular the mean differences in bone age were similar to reported interobserver variability. This has led to our centre finding automated bone age software acceptable for use locally.