Session P95.2

3D Cardiac MRI Data Visualization Based on Volume Data Preprocessing and Transfer Function Design

F Yang, WM Zuo, KQ Wang*

Harbin Institute of Technology
Harbin, China

The variety and complexity of cardiac tissue makes it very challenging to visualize 3D cardiac magnetic resonance image (MRI) data. Classical transfer function, one of the key technologies in direct volume rendering, usually suffers from the problem of data quality degradation and structure complexity of the volume data, resulting in the poor quality of the volume-rendered image. In this paper, we proposed an adaptive volume data preprocessing approach, and used a semi-automated transfer function for volume rendering of the pre-processed cardiac MRI data. To improve the quality of 3D cardiac MRI data, we first used a three-dimensional median filtering method for data denoising, and then proposed an adaptive ellipsoidal Gaussian filtering scheme for local-feature-preserving data smoothing. To determine the parameters of the ellipsoidal Gaussian filter, we first compute the first and the second derivative of a specific voxel to decide its distance to the boundary and its gradient direction. If the voxel is close to the boundary, we will design an ellipsoidal Gaussian filter along the gradient direction; else we will use a circular 3D Gaussian filter for image smoothing. In such a way, we can perform volume data smoothing without losing the detailed information of the material boundary. For effective visualization, we implemented a volume rendering pipeline with ray casting and the semi -automated transfer function. To design an appropriate transfer function, we first calculate the first and the second derivatives of each voxel, and use all of them to construct a histogram volume to capture the relationship of the distance to boundary with the value, the first and the second derivatives. Using the histogram volume, the transfer function is designed by mapping data value and the first derivative of each voxel to its position to the boundary, and a semi-automatic transfer function design scheme is adopted to allow user create transfer function to visualize the information which he is interested in. Finally, the efficiency of the proposed 3D cardiac MRI data visualization method is verified using the MRI data of a sheep heart.

(Abstract Control Number: 269)