Technical note: Standardized and semiautomated Harris lines detection

• Published in: American journal of physical anthropology Vol. 137; no. 3; pp. 362 - 366
• Main Authors: Suter, S., Harders, M., Papageorgopoulou, C., Kuhn, G., Székely, G., Rühli, F.J.
• Format: Journal Article
• Language: English
• Published: Hoboken Wiley Subscription Services, Inc., A Wiley Company 01.11.2008; Wiley-Liss
• Subjects: Adolescent; Adult; Age Determination by Skeleton - methods; Age Factors; algorithm; Biological and medical sciences; Body; bone growth; Child; feature detection; Female; Human remains; Humans; Image Processing, Computer-Assisted; Investigative techniques, diagnostic techniques (general aspects); Male; Medical sciences; Middle Aged; Osteoarticular system. Muscles; Paleopathology; Physical anthropology; Radiodiagnosis. Nmr imagery. Nmr spectrometry; radiograph; Software; Tibia - anatomy & histology; Tibia - diagnostic imaging; Tibia - growth & development; Human; radiograph; Growth; bone growth; Algorithm; Semiautomatic; Osteoarticular system; Radiography; feature detection; Image analysis; paleopathology; Bone; Detection
• ISSN: 0002-9483; 1096-8644
• DOI: 10.1002/ajpa.20901

Arrest in long bone growth and the subsequent resumption of growth may be visible as radiopaque transverse lines in radiographs (Harris lines, HL; Harris, HA. 1933. Bone growth in health and disease. London: Oxford University Press). The assessment of individual age at occurrence of such lines, as part of paleopathological skeletal studies, is time‐consuming and shows large intra‐ and interobserver variability. Thus, a standardized, automated detection algorithm would help to increase the validity of such paleopathological research. We present an image analysis application facilitating automatic detection of HL. On the basis of established age calculation methods, the individual age‐at‐formation can be automatically assessed with the tool presented. Additional user input to confirm the automatic result is possible via an intuitive graphical user interface. Automated detection of HL from digital radiographs of a sample of late Medieval Swiss tibiae was compared to the consensus of manual assessment by two blinded expert observers. The intra‐ and interobserver variability was high. The quality of the observer result improved when standardized detection criteria were defined and applied. The newly developed algorithm detected two‐thirds of the HL that were identified as consensus lines between the observers. It was, however, necessary to validate the last one‐third by manual editing. The lack of a large test series must be noted. The application is freely available for further testing by any interested researcher. Am J Phys Anthropol, 2008. © 2008 Wiley‐Liss, Inc.