Available at: http://www.lems.brown.edu/~msj/cs292/assign3/volumeVis.html
 

Introduction:  First and foremost medical imaging as a field is concerned with visualizing the internals of the body in order to diagnose, treat,  and understand the human body; diseased and healthy.  Image data comes in many forms, some more efficacious then others for visualization, some better suited for computation.  Some standard techniques for visualization and computation are examined here.

Medical imaging targets are often 3D entities, but many imaging modalities extract data in 2D slices.  The result is a discrete stack of 2D images to represent a 3D object.  These 2D slices can be viewed three different ways, a view corresponding to slicing alone one of the three axes.  These views are know as axial, coronal, and sagital.

Axial: Slices are in the XY plane.
Coronal: Slices are in the XZ plane.
Sagital: Slices are in the YZ plane.

Figures 1,2 and 3 show the slices along the three axies from stacked data of the skull.
 
 

Matlab Code
slice3d.m
scale.m

A simplistic method to extract a surface model of a structure of interest from the 3D volumetric image is to pick an appropriate intensity value as the isovalue and reconstruct the corresponding isosurface.  Figures 4,5 and 6 show isosurfaces at several isovalues and angles of the head,spine and DT data.
 

Computation is often critical to medical imaging applications.  Calculation of normals is a standard procedure in understanding and enabling techniques in surface description, registeration, and diagnosis.  The normal of a 3D surface represented as the zero level-set of its distance transform can be calculated by the partial derivatives at each grid location.  The partial derivative will approximated by taking the central difference.  Figure 7 shows different views of the DT surface with the calculated normals applied.

Matlab Code
displayNorms.m
calculateNormals.m

Many medical imaging modalities generate image data in stacks of 2D image slices.  Often times it is desirable to view these slices as a 3D object possessing a surface model.  Furthermore, many computational techniques require a surface model.  A standard technique of building a 3D surface model from 2D segmented image slices is the Signed Distance Transform.  Figure 8 shows 21 hand segmented 2D slices. Figure 9 shows the 3D surface constructed from the distance transform of the slices.
 

Matlab Code and Data
segSpine.m
myspine.mat
 

Support Code taken from Medical Imaging Page
displayDT.m
displayIso.m