Multi-phase simultaneous segmentation of tumor in lung 4D-CT data with context information

Abstract

Lung 4D computed tomography (4D-CT) plays an important role in high-precision radiotherapy because it characterizes respiratory motion, which is crucial for accurate target definition. However, the manual segmentation of a lung tumor is a heavy workload for doctors because of the large number of lung 4D-CT data slices. Meanwhile, tumor segmentation is still a notoriously challenging problem in computer-aided diagnosis. In this paper, we propose a new method based on an improved graph cut algorithm with context information constraint to find a convenient and robust approach of lung 4D-CT tumor segmentation. We combine all phases of the lung 4D-CT into a global graph, and construct a global energy function accordingly. The sub-graph is first constructed for each phase. A context cost term is enforced to achieve segmentation results in every phase by adding a context constraint between neighboring phases. A global energy function is finally constructed by combining all cost terms. The optimization is achieved by solving a max-flow/min-cut problem, which leads to simultaneous and robust segmentation of the tumor in all the lung 4D-CT phases. The effectiveness of our approach is validated through experiments on 10 different lung 4D-CT cases. The comparison with the graph cut without context constraint, the level set method and the graph cut with star shape prior demonstrates that the proposed method obtains more accurate and robust segmentation results.

Lung tumor segmentation in 4D CT images using motion convolutional neural networks

“…[38][39][40][41][42] A previous study achieved the DSC between manual and automatic segmentation on test datasets to be 0.855 ± 0.048, 0.763 ± 0.072, and 0.831 ± 0.068 for graph cut algorithm,level set method,and the graph cut with star shape prior, respectively. 39 Speight et al 41 achieved segmentation accuracy with DSC = 0.856 ± 0.045 by using Bspline method. Gaede et al 42 achieved a mean DSC of 0.80 between automatic and manual segmentation processes.…”

“…In a study by Shen et al, physicians had to select object and background seeds in the image of one phase to begin automatic contouring. 39 It is also important to note that VoxelMorph required segmentation of tumor in one phase in order to bring tumors in remaining phases to this one. In contrast, our proposed network requires neither the first frame's tumor annotation nor derived features to learn the differences amongst different labels, allowing the fully automatic segmentation of lung tumor on 4D CT.…”

“…Although several ML methods have been proposed for tumor tracking, 43,44 they require users to manually extract features to be used in the mathematical models. In a study by Shen et al., physicians had to select object and background seeds in the image of one phase to begin automatic contouring 39 . It is also important to note that VoxelMorph required segmentation of tumor in one phase in order to bring tumors in remaining phases to this one.…”

“…Beside DL‐based methods, different iterative algorithms have been reported in the literature for automatic segmentation of lung tumors 38–42 . A previous study achieved the DSC between manual and automatic segmentation on test datasets to be 0.855 ± 0.048, 0.763 ± 0.072, and 0.831 ± 0.068 for graph cut algorithm, level set method, and the graph cut with star shape prior, respectively 39 . Speight et al 41 .…”

Multilayer Joint Segmentation Using MRF and Graph Cuts

“…Considering several images at once for segmentation allows for an increase of the information while rising new questions on its exploitation. Applications of this problem include (but are not limited to) video object-based segmentation [20], interactive image segmentation [17,35], watermark removal [13] and 3D reconstruction [3].…”