The approach that was implemented was based on edge detection, line tracing, and histogram thresholding techniques . The requirements for this process do not differ significantly from those followed in standard chest radiography (CXR) and several of the concepts described in CXR literature are applicable to CT as well . One primary issue in this module was the desired level of accuracy in the removal of the external signals, i.e., signals from the rib cage and spine. Increasing the accuracy level, increased the computational requirements and the complexity of the methodology. Figures 4.11 and 4.12 show the external signal removal for the slices of Figs. 4.6(a) and 4.7(a).
A histogram equalization approach was used to remove the regions that correspond to the rib cage and spine that usually are the highest intensity regions in the image. Points on the rib cage were defined using the pixel characteristics of the rib cage and these points were interpolated using a spline interpolation technique . The boundary of the rib cage was then estimated and removed.
Our enhancement approach aimed at increasing the image contrast between the pancreas and organs in close proximity. A histogram equalization approach was implemented for this purpose and yielded satisfactory results (Gaussian and Wiener filters seemed to benefit these images as well) . Wavelet-based enhancement was also considered as an alternative option for removing unwanted background information and better isolating the signals of interest . The method was promising but may present an issue when used in combination with registration or reconstruction processes.
Enhancement generally benefits CAD algorithms but in 3-D imaging modalities like CT, it may have an adverse effect on the registration of the 2-D data, if it is not uniformly done across slices. Wavelet-based enhancement may worsen
the situation since it operates in the frequency domain and may not necessarily preserve the spatial features of the CT images as needed for registration. A standardization or normalization method may offer a solution in regaining all spatial information when transforming from the frequency back to the spatial domain. However, no such method was established for this application or is readily available. If registration is not part of the process, the enhancement step could significantly benefit subsequent clustering and classification on the CT images .
Was this article helpful?