Evaluation of 3D gradient filters for estimation of the surface orientation in CTC

The extraction of the gradient information from 3D surfaces plays an important role for many applications including 3D graphics and medical imaging. The extraction of the 3D gradient information is performed by filtering the input data with high pass filters that are typically implemented using 3×3×3 masks. Since these filters extract the gradient information in small neighborhood, the estimated gradient information will be very sensitive to image noise. The development of a 3D gradient operator that is robust to image noise is particularly important since the medical datasets are characterized by a relatively low signal to noise ratio. The aim of this paper is to detail the implementation of an optimized 3D gradient operator that is applied to sample the local curvature of the colon wall in CT data and its influence on the overall performance of our CAD-CTC method. The developed 3D gradient operator has been applied to extract the local curvature of the colon wall in a large number CT datasets captured with different radiation doses and the experimental results are presented and discussed