Purpose/Objective: Re-contouring regions of interest (target or sensitive structures) on each stage of treatment image is the one of the most challenging tasks in 4-D planning and adaptive radiotherapy when multi-sets of image data are available. The requirements of manually re-contouring prevent 4-D planning and plan modification being done routinely. An automatic re-contouring technique is presented, which could reduce or even eliminate such obstacles. Materials/Methods: We use CT images of lung patients for testing. The CT of inhale stage is used as reference image and CT of the other stages are used as test image. This technique consists of four steps. 1. Contouring and surface reconstruction: The regions of interest are contoured manually or semi-automatically in the reference image . The reference surface are built based on these contours using a triangle surface reconstruction algorithm. 2. Deformable registration: An in house fast automatic deformable registration technique is developed. A 3-D displacement map is calculated based on the deformable registration of the reference and test images. 3. Surface deformation: The vertices of the reference surface are displaced in accordance with the displacement map, which results in the test surface. 4. Re-contouring based on deformed surface: The test contours are reconstructed by plane-intersecting the test surface slice by slice. Results: The left panel of Figure 1 shows the surfaces of lung and spinal cord that are reconstructed from the reference CT image. The right panel shows the results of surface deformation based on deformable registration. Two slices of test image and their contouring results are illustrated in Figure 2, where solid lines show the results of automatic contouring, and dashed lines show their reference correspondences. The results show that this technique re-contours the regions with large deformation (such as lung) or large misalignment (such as lung and spinal cord) quite accurately. Conclusions: Deformable registration of two stages of the medical image is an essential technique for 4-D planning and adaptive radiotherapy, because it is the basis for cumulative dose calculation. Surface reconstruction is a valuable technique for viewing and manipulating 3-D object. Through the combination the deformable registration and surface reconstruction, automatic re-contouring of ROIs becomes feasible. This technique also provides a metric for evaluating the deformable registration algorithm.
Lu, W,
Olivera, G,
Zhang, T,
Chen, M,
Ruchala, K,
Automatic Re-Contouring for 4-D Planning and Adaptive Radiotherapy. Radiological Society of North America 2004 Scientific Assembly and Annual Meeting, November 28 - December 3, 2004 ,Chicago IL.
http://archive.rsna.org/2004/4417907.html
