Qualitative femur bone tensile strength analysis in radiographic images using wavelets

In this work, the tensile strength of femur bone is qualitatively analyzed using radiographic images. The sub anatomic tensile regions were delineated from conventional planar radiographic femur images using digital image processing algorithms. The normal and abnormal images are then subjected to Daubechies5, coiflet5 and Haar wavelets decomposed at three levels to derive approximation and detail coefficients. The qualitative analyses were also performed on the delineated images to derive apparent mineralization and total area. The values of higher order energy parameter are derived for both approximation and detail coefficients in each level of decomposition and are correlated with apparent mineralization for analysis. Results show that the low values of energy correlates well with abnormalities in all levels of decomposition for all the three wavelets. Among all, the values of energy derived out of approximation coefficient using Haar wavelets show higher degrees of correlations in both primary tensile and secondary tensile regional strength. The variations between normal and abnormal were also found to be statistically significant. Poor correlation was observed in the case of detailed coefficients for all the wavelets irrespective of levels. Among all wavelets, the parameters decomposed at level one approximation coefficient of Haar wavelet appear to be a useful predictor for classifying normal and abnormal samples. This could be attributed to sensitivity of adopted wavelets to the architectural changes and discontinuities in regional distribution of trabecular pattern in tensile region. It appears that this methodology could be used for gross abnormality detection, micro-damage studies and modeling the mechanics of soft tissue in diseases. Thus wavelet extracted feature on tensile trabeculae regions of radiographic femur images could be a used as an index for automated screening of bone strength.