Principal Investigator Reiner Wilhelms
A new finite element model is used to analyze the anatomy of the human tongue. The original data consist of axial sections of a frozen cadaver taken 1/3 mm apart. A portion of the cryosection data of the female cadaver was used, the oral area were cropped out and combined into a stack of RGB images. Using C programming and the Matlab software for display, a software was built that allows to generate and edit small rectangles representing sections through the data (voxel block). Rectangular regions are sampled in a C-written renderer and combined into a view. A small feedforward neural network had been used to distinguish between tissue and the blue colored filler in which the cadaver was embedded. This idea can to some extent be applied to differentiate between various tissue components. For each tissue, such as blood lumen, dark musculature, "red" musculature, fat, bone, teeth, etc. a feed-forward network with a low number of hidden nodes was trained. The trained network can then compute the likelyhood of an RGB sample vector being part of a particular tissue.