Modified Non-Gaussian Models with Implicit Functions for Pulmonary Nodule Simulation

Background and Purpose: The simulation of pulmonary nodule is facilitating the studies of lung nodule detection. To overcome the shortcoming of traditional Gaussian models that can only generate the sphere and ellipsoidal nodules, an effective approach with implicit functions is proposed and developed in this article. Methods: The traditional ternary function model is improved for the simulation of different kinds of nodules with various shapes. The realization process has two main stages. Firstly, the shape modeling process for the simulation of the sphere and ellipsoidal nodules is provided. Secondly, the traditional ternary function is modified by related adjustment factors for the simulation of nodules with irregular, halo and other shapes. To verify the effectiveness and feasibility of the presented model, the virtual nodules were synthesized and inserted into the computed tomography (CT) images, then four observers tried to distinguish the simulated nodules from real nodules. Results and Conclusions: A novel and robust model based on modified implicit functions is proposed and developed for the simulation of pulmonary nodules on chest CT images. Four types and eight kinds of lung nodules were simulated by the proposed model, especially the irregular and halo nodules. In addition, nodules with other shapes can be created by changing parameters of the presented model. The experimental results show that the proposed approach outperformed the conventional method. The presented approach has potential application in areas such as nodule size measurement, tumor growth rate assessment, training and testing of machine learning algorithm in nodule detection, and other tasks.