Lymph node segmentation on CT images by a shape model guided deformable surface methodh

Complete fully automatic model-based segmentation of normal and pathological lymph nodes in CT data

“…Enlarged, necrotic, or fuzzybounded lymph nodes were not addressed, which is however very important for tumor diagnosis. A similar approach was presented by Maleike et al [11], who used a statistical instead of a Stable Mass-Spring Model. This statistical model uses very few shape modes to restrict the model to rather elliptical shapes.…”

Graph-Based Segmentation of Lymph Nodes in CT Data

“…The author reported that the method could detect 57.0% of the enlarged lymph nodes with approximately 58 false positives per data sets. In [24], a deformable surface search was used in combination of statistical shape model for lymph node segmentation. Additionally, an application which allows manual interactive correction of errors was provided to help the algorithm in converging to the desired object contours.…”

Preoperative workflow for lymph nodes staging

“…In the work of Kitasaka [10] a 3D minimum directional filter is employed for enhancing blob structures, and then a region growing algorithm is applied in order to cover whole lymph nodes, since detected regions are just the center of the detected structures. Some other researches have proposed the application of prior information on the shape of the lymph nodes: while Slabaugh and Unal propose a graph cut approach and the use of an elliptical shape prior to constrain the segmentation [18] or in an other work, after the definition of spatial and intensity constraints, they estimate the initial contour and approximate and propagate it as an ellipse toward the lymph node boundaries [21], Maleike et al [14] have developed an algorithm based on a deformable surface search, which uses statistical shape models to restrict free deformation, and have constructed an ellipsoid shape model, which strives for a surface with strong gradients and userdefined gray values. Other proposed approaches consist in the application of active contours [7,23] or in the concurrent use of different information such as the grey value range, directed contour information as well as shape knowledge [4] or in the development of atlas-based delineation algorithms [3].…”