Interactive visualization of three-dimensional confocal microscopy data

Qiqi Wang, Purdue University

Abstract

Confocal microscopy is a powerful tool for measuring 3D cellular structures and behaviors. However, the measured data have significant noise due to light absorption and scattering. Such noise causes intensity attenuation, making data analysis more challenging. Besides, confocal datasets have additional distinct properties, including thin data volume, multiple channels, and object diversity. These properties have not been considered in the previous visualization techniques, and the resulting visualization is not satisfactory for researchers and practitioners in biology. This thesis develops new methods to visualize confocal data with consideration of the distinct properties of the data. The new contributions are: 1. An independent model of energy emission and attenuation is proposed. The physical basis is that the energy generation and attenuation involve different aspects of the material. This model is not only physically correct, but achieves better visualization than the commonly used locked model. 2. Interactive volume rendering is implemented using 3D textures. The stencil approach is proposed to combine texture blending along the three principal axes. This approach has eliminated the drawback of the previous approach. 3. In confocal data, since multi-channel datasets are precisely registered and may be very different in different channels, this thesis proposes two-channel classification to distinguish objects in the data. 4. Finally, our visualization program offers an option of interactive shading. The implementation is made based on the advanced graphics hardware and the OpenGL Shading Language (GLSL).

Degree

Ph.D.

Advisors

Sun, Purdue University.

Subject Area

Cellular biology|Computer science

COinS