Fully automatic segmentation of phalanges from hand radiographs for bone age assessment

Abstract

Segmentation of bones from hand radiograph is an important step in automated bone age assessment (ABAA) system. Main challenges in the segmentation of bones are the intensity inhomogeneity caused by the irregular distribution of X-rays and the overlapping pixel intensities between the bone and soft tissue. Hence, there is a need to develop a robust segmentation technique to tackle the problems associated with the hand radiographs. This paper proposes a fully automatic technique for segmentation of phalanges from left-hand radiograph for bone age assessment. The proposed technique is divided into five stages which are pre-processing, extraction of Phalangeal region of interest, edge preservation, segmentation of phalanges and post-processing. Quantitative and qualitative results of proposed segmentation technique are evaluated and compared with other state-of-the-art segmentation methods. Qualitative results of proposed segmentation technique are also validated by different medical experts. The segmentation accuracy achieved by proposed segmentation technique is 94%. The proposed technique can be used for development of fully ABAA of a person for better accuracy.

Keywords:

• Hand bone segmentation
• bone age assessment
• level set function
• Phalangeal region of interest (PROI)

Acknowledgements

The authors would like to thank the editor and anonymous reviewers for their constructive comments, which improved the quality of the manuscript. The authors would also like to thank Dr. Pranav Nagarsenkar from Goa Medical College, Goa, India and Dr. Amresh Naik from Directorate of Health Services, Panjim, Goa, India for helping with creation the of ground truth images, validating the segmentation results and reviewing the manuscript. Special thanks to Professor Roberto A. Lotufo from the University of Campinas for helping out with the license of morphology toolbox, which has been an immense help.

Notes

Supplemental data for this article can be accessed https://doi.org/10.1080/21681163.2017.1416491.

2 No potential conflict of interest was reported by the authors.