Research On Three Dimensional Direct Volume Rendering Of Ultrasound Heart Images

Traditional three-dimensional visualization methods used in ultrasound data set lied on pre-segmentation and noise reduction, then series of patches were used to form a mesh based on the surface rendering. However, these methods depend on the success of pre-segmentation heavily. With the development of Direct Volume Rendering (DVR) technique, traditional Volume Rendering was utilized in three-dimensional visualization of ultrasound data. However, the traditional Volume Rendering algorithm by Levoy mainly aimed at visualization of CT and MRI datasets, lacking consideration of Speckle noises. Thus, how to built a robust opacity transfer function (OTF) that can suppress noise while enhance the rendered image contrast effectively become the key to the application of Direct Volume Rendering in three-dimensional visualization of ultrasound data.The appearance of speckle noise, the lower gray dynamic range together with quite wide blurring area all make the three dimensional visualization of ultrasound data difficult. In the thesis, we find that direct volume rendering technology did not consider the noise interference in its physical model, but a powerful rendering function can realize the setting of optical attributes in voxel, so that the interference of Speckle noise can be suppressed effectively by controlling OTF. Based on the DVR method, the thesis investigated voxel brightness change direction in different coordinate system. We changed the traditional OTF by adding voxel brightness change direction into OTF, thus we built a three dimensional OTF combined with the voxel color, voxel gratitude magnitude and voxel brightness change direction.Different opacity transfer functions were studied in the thesis. Piecewise Linear function and parabolic function were compared. Through experiment, we find parabolic opacity transfer function is more effective in enhancing the rendered image contrast. In order to verify, a simplified volumetric composition process was built and used to compare the two functions, using the depth contrast and tissue contrast. Finally, the effete which the two functions do to the final rendered result were compared and analyzed.In order to verify the effectiveness of the new OTF in the thesis, we built a virtual dataset contain different gray levels and all burred by speckle noise. The visualization result using the proposed method and the traditional Shear-Warp algorithm were compared, it can be concluded that our method perform more effective in speckle suppression. In the thesis, dynamic ultrasound human heart data was visualized using the new OTF. The three dimensional visualization results by both traditional Shear-Warp algorithm and the modified algorithm which adopts new OTF were compared and analyzed.Based on the visualization results of ultrasound datasets, we can conclude that the OTF plays an important role in the application of DVR in the ultrasound dataset. The result indicates that during the visualization process the proposed OTF can overcome the interference of speckle noise to some extent. Meanwhile, clearer anatomical structure of human heart with enhanced image contrast can be observed from the result. The thesis targeted mainly on improving the key part of DVR, the OTF, thus changed the traditional OTF which lies heavily on pre-procession of the dataset. Comparatively, the proposed OTF offers better classification and opacity mapping while overcome the interference of the speckle noise before composition.