Existing volume visualization algorithms lack sophisticated volumetric modeling capabilities to represent, construct and manipulate objects of interest within a volumetric dataset. Integrating manipulations such as non-linear deformations and constructions similar to those in solid modeling into the visualization process would have significant value in surgical planning and simulation. Interrogative visualization-the ability to extract, enumerate, deform and track objects of interest is also crucial in other visualization tasks. The conventional approach to visualize the volume deformation and construction process is to construct an intermediate volume for each step of manipulation. This is prohibitively time consuming and costly. This paper presents an approach to overcome this constraint. We model the target capture space as octrees. The target octree is adaptively subdivided to efficiently balance the complexity of deformation and visualization. The source volumetric data can either be in octree or conventional 3D array format. Deformations are specified by landmark displacements, and in-between deformation is filled in by scattered data interpolation. Volumetric modeling is supported by a Constructive Solid Geometry (CSG) structure. Spatial Coherency is exploited in the view-dependent node traversal and projection of individual octree nodes.