International Congress Series 1281 (2005) 1380
An initial approach to deformation modeling for soft tissue Chen Hongjuna, Shao Fanb, Ng Wan Singa,*, Shi Damingc a
Computer Integrated Medical Intervention Lab., School of Mechanical and Aerospace Engineering, Nanyang Technological University, Singapore b School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore c School of Computer Engineering, Nanyang Technological University, Singapore
Keywords: PCA; Correspondence; Spherical Harmonics; Soft tissue modeling
1. Introduction There exist 3 classes of approaches to soft tissue modeling: non-physical, physical and statistical modeling. This paper follows the statistical approach. 2. Methods In [1], the patient’s anatomical shape vector s and deformation vector q are concatenated into x, which is used as training sample for a PCA-based statistical model. To eliminate the requirement for intra-sample correspondence, this paper proposes to use the deformed and un-deformed shape vectors to form the training samples. To speed the model training process, the combination of this statistical model with Spherical Harmonics (SH) [2] representation method in frequency domain is proposed. 3. Results and conclusions In the experiments, 30 samples generated from ANSYS were used. The results are listed in Table 1, which demonstrates that the adjusted algorithms can successfully solve the intra-sample correspondence problem and can effectively decrease computation time. References [1] C. Davatzikos, D.G. Shen, A. Mohamed, A framework for predictive modeling of anatomical deformations, IEEE Trans. Med. Imag. 20 (2001) 836 – 843. [2] M. Kazhdan, T. Funkhouser, S. Rusinkiewicz, Rotation invariant spherical harmonic representation of 3D shape descriptors, Proc. of the Eurographic/ACM, SIGGRAPH Sym. on Geometry Processing, vol. 6, 2003, pp. 156 – 164. Table 1 Relative prediction error and model training time Experiment condition
Space domain
Frequency domain
PCA-based algorithm
With intra-corr. Without intra-corr. a
Adjusted algorithm
Adjusted algorithm with SH
Error
Time (s)
Error
Time (s)
Error
Time (s)
3.70% 343.40%
609.8 615.9
3.40% 4.10%
615.5 620.9
N/A 4.20%
N/A 354.4a
Including 330.0 s used for computing SH coefficients; based on Dell workstation, P4 2.4G, 512M.
T Corresponding author. E-mail address:
[email protected] (W.S. Ng). 0531-5131/ D 2005 CARS & Elsevier B.V. All rights reserved. doi:10.1016/j.ics.2005.03.272