Automated extraction of the endorrhachis from MR lumbar images by fuzzy inference techniques

Surface rendering of the endorrhachis is one of the effective techniques to diagnose diseases of the endorrhachis caused by spine malformations. It requires a procedure for segmenting the 3D endorrhachis from many MR images consisting of the whole human body, however, few automated extraction methods have been developed. We propose a fuzzy rule-based approach for extracting the endorrhachis from MR lumbar images. Our target portion, the endorrhachis, consists of the spinal cord and the nerve roots. This method is based on knowledge of location and knowledge of intensity. They are expressed by fuzzy if-then rules and compiled to a total degree as the measure of extraction. Our extraction method can be divided into two steps. Firstly, the method extracts the spinal cord by using thresholding and labeling. Secondly, our method extracts the nerve roots with the knowledge derived from anatomical information. The experimental results showed that our method successfully extracted the target portion, and that the results were useful to diagnose the compression of the endorrhachis.