The aim of this thesis is to enable the automated evaluation of scintigraphic images of the cervical spine of horses using a CNN. We want to detect inflammation in the facet joint C6/C7. Veterinarians assess this disease on the basis of the radiation intensity and its comparison with the radiation intensity of the facet joint C3/C4, as the radiation intensity alone varies greatly from individual to individual. Measuring these differences in intensity is very complicated and time-consuming. An automated evaluation would therefore be highly desirable.
First, the corresponding facet joints must be masked in the images. These images should then be used to train a CNN, which should learn to classify the images into “sick” and “healthy”.
The data is available in Dicom format.

key words: image processing, machine learning, deep learning, classification, dicom, masking