Automated particle counting with applications to neurology
In this thesis an algorithm is developed to count regular shaped particles in digitized images. The algorithm was applied to counting axons in nerve fasciculi regions.
Images of the data are digitized using a special Tv camera. The digitized data is then preprocessed before sent to the recognizer.
Particles are located in the digitized image by matching them with a representative mask. Closeness of the match is determined using binary correlation.
After the slides of the fasciculi regions have been digitized, the system runs completely automatically, taking in data from magnetic tape and storing the results on magnetic disks. Using manually counted regions as a basis for comparison the system is more than 90 percent accurate for the three nerves used.
This algorithm shows promise as a useful tool in any application where multiple particles are counted. As illustrated in this thesis one direct application is in the field of neurology.
Thesis77.C283.pdf
5.11 MB
Adobe PDF
de27a1ec7ac6cb1c70852405f4eb897a