"Cryosectioning" is a rather gruesome, but scientifically fascinating process that starts by freezing a cadaver. The cadaver is then cut into very thin slices and each slice photographed digitally at high resolution. The resulting images are stacked within a 3D coordinate space to create a volume data set that depicts the cadaver. Image processing and transfer functions applied to the data turn selected tissues transparent to reveal inner structure.
These images are based upon cryosections of a human brain. Transfer functions start by turning transparent everything outside of the brain. Volume rendered, the result is a realistic model of the brain. The data can be cut any way you like to look inside. All of this can form the basis of automatic or semi-automatic "segmentation" that labels different regions of the brain. Once segmented, calculations can be made for the volume of different brain regions of the brain. Repeat this for a large selection of normal and dysfunctional brains and, it is hoped, a pattern emerges that can lead to a better understanding of what makes a brain work, or not work. And this could lead to ways to diagnose problems early... and of course without slicing up the patient's brain.
This work was funded by the National Institutes of Health. Development was in C++.
||Head cryosections and CT scans|
Medical imaging has several different ways to look inside a patient. While cryosectioning is only used on those that have passed on (thank goodness), CT and MRI scans can be used on the living. Each technique has its strengths and weaknesses. Combining them can use the best of each. For instance, a CT scan clearly shows bone. Cryosectioning shows true color, but not all things that are white are necessarily bone. So, by combining cryosectioning and CT scans we can use CT data to identify bones, extract them, and reveal cryosectioning data for the tissue around the bone (skin, muscle, brain, etc.).
These images combine cryosectioning and CT scan data from the Visible Male Human data set from the National Library of Medicine. Volumetric scene graphs combine the data and apply transfer functions to set opacity and color. More interestingly, the scene graphs treat the volumes as "shapes" that can be translated, rotated, scaled, sliced, and composited to form a volumetric scene. This lets us slice open the skull and rotated it out of the way to reveal the brain inside. Or divide the skull in two and pull it apart to show the brain, then slice the brain. And so on. The approach enables a much more flexible manipulation of volumetric data than is possible using traditional volume rendering.
This work was funded by the National Science Foundation. Development was in C++.