A model retrieving based method for bolus shaping

摘要

Bolus is a sheet of material commonly used in the treatment of superficial tumors for desired dose distributions. Existing methods of the bolus shaping cannot meet required accuracy to cover some irregular surfaces. This paper introduces a shape retrieving method to increase the bolus accuracy and process efficiency. Common human surfaces that need bolus in the treatment are pre-processed by segmentation based on the surface flattenability and deformation. The segmented surfaces are unfolded to form 2D shapes with the minimal deformation and saved in a model base. A bolus can then be quickly formed by retrieving the matched bolus model in the model base using the patient data captured by a Kinect motion sensor. To match the model in a high accuracy, features of patient’s data are first extracted using the Laplacian matrix to build a feature space. The features are matched using an iterative closest point (ICP) method. An example of the human nose bolus is presented to show the proposed method.