In the past, rapid volume rendering of large fluorescence microscopy data sets required expensive 3D workstations or personal computers equipped with voxel coprocessor boards. Near real-time rendering of such data is now possible using personal computers equipped with low-cost ($150-400) video boards that employ the latest 3D graphics processors. We have developed a volume rendering program (Voxx) that uses such processors to render confocal and two-photon microscopy data [Clendenon, et al., Am. J. Physiol. Cell Physiol. 282: C213-C218, 2002]. However, the typically noisy, low-contrast nature of such data can make it difficult to produce consistently high-quality 3D effects.

Here we describe our recent experiments in which tensor-based techniques were used to produce improved 3D images. Specifically, we compute the second-order symmetric tensor for a 3D filter window centered on each element in the voxel array. The eigenvectors extracted from the tensor matrix are used as estimates of the orientation of fluorescing structures within each window, and shape factors are derived from the eigenvalues. The shape and orientation data can be used in adaptive filters, and in lighting and shading calculations during volume rendering. Such techniques have recently been used in volume rendering of diffusion tensor MRI data [Kindlmann, et al., IEEE Trans. on Vis. and Comp. Graph. 6: 124-138, 2000], and our results demonstrate that they can also be useful in 3D fluorescence microscopy.

J.L. Clendenon, J.M. Byars, C.L. Phillips, K.W, Dunn, "Improved volume rendering techniques for confocal and multi-photon microscopy“ (abstract), Molecular Biology of the Cell, vol. 13 supplement (November, 2002), p. 551a. poster at the 42nd Annual Meeting of the American Society for Cell Biology