296	Q. Zhang et al.

Fig. 13.5 Volume rendered images of volumetric human cardiac CT data set

major diagnostic criteria that must be considered. Even though DVR generally has high sensitivity and specificity for diagnosis [28], it is computationally intensive, so interactive performance is not always feasible. Another disadvantage is that it may be difficult to interpret the “cloudy” interiors that can result from the raytracing process. For detailed observation of specific lesions, slab imaging, where thick slices are rendered with DVR or MIP, has generally been used in clinical diagnosis [29]. In addition, the combination of cross-sectional MPR and DVR can significantly increase the interpretation rate of anatomical structures [30]. DVR has a wide range of clinical applications. For example, to perform renal donor evaluation, Fishman et al. [31] used DVR and MIP to display the renal vein of CT angiography (CTA) data, as shown in Fig. 13.6, for the DVR generated image, the left gonadal vein (large arrow) is well defined (a), while the locations of the renal vein and gonadal vein (arrow) are inaccurately depicted in the MIP generated image (b). Gill et al. [32] used DVR and MinIP to show the central airway and vascular structures of a 60-year-old man who underwent double lung transplantation for idiopathic pulmonary fibrosis, evaluating posttreatment in a noninvasive manner (Fig. 13.6c,d).

13.3 Volume Rendering Principles

The core component of DVR is to solve the volume rendering integral that describes the optical model. Different DVR techniques share similar components in the rendering pipeline, the main difference being the order in which they are applied, and the manner in which they traverse the volumetric data. Due to the rapid development of programmable GPUs, many CPU-based algorithms and techniques have been or can be implemented on GPUs. In this paper, for the texture-mapping based DVR, we refer to algorithms that use a fixed graphics pipeline, while for GPU-based raycasting, we refer to DVR implemented on GPUs.

13 Medical Image Volumetric Visualization: Algorithms, Pipelines...

297

Fig. 13.6 DVR, MIP, and MinIP applications: (a) Coronal oblique DVR image of kidney and veins; (b) MIP display of the same data as (a) [31]. (c) DVR generated image of the central airway and vascular structures. (d) Coronal MinIP image of the same data as (c) [32]

13.3.1 Optical Models

The volume rendering integral is still often based on a model developed by Blinn [33] describing a statistical simulation of light passing through, and being reflected by, clouds of similar small particles. The optical models may be based on emission or absorption individually, or both, depending on the applications [34]. To reduce the computational cost, Blinn assumed that the volume is in a low albedo environment, in which multiple reflections and scattering of the particles are negligible. In this case, a light emission–absorption model is an optimal balance between realism and computational complexity, where every particle absorbs incoming light and emits light on its own without scattering between particles other than in the viewing ray direction [16]. Equation (13.1) demonstrates this procedure.