| Abstrakt: |
Human hand Bone Age Assessment (BAA) is commonly used by paediatrics for diagnosing the childś health, crime investigation, dead human identification (in case of natural disasters) etc. BAA can be estimated using Epiphysis and Carpal bones of hand X-ray images. These X-ray images may be prone to noise, due to which it might affect the process of background separation and auto-segmentation while performing BAA. Methods existing in the literature of BAA lack in resolving one or more issues (noise reduction, background separation and bone pixel segmentation) individually. This paper proposes a new method to segment Epiphysis bones using the (i) Wavelet packet transformation for noise suppression, (ii) Background suppression using texture features of the image, (iii) Enhancement of bone pixels using histogram equalization and (iv) Finally segmenting the Epiphysis region of interest using clustering method. Performance analysis is performed using two quantitative evaluation methods i.e. supervised and unsupervised evaluation. Unsupervised approach uses Mean Structure Similarity Index (MSSIM) and Homogeneity. Supervised approach uses Precision, Recall, Sensitivity, Accuracy and Error Rate. Proposed method for segmenting the epiphysis bones has shown better accuracy rate of 0.9701. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink