Segmentation of CT Brain Data

Image segmentation is the process of assigning labels to individual pixels in a volumetric data set, based on some criteria in the image. In medical image segmentation, pixels are labeled by tissue type. Image segmentation is required in many medical applications ranging from the education and assessment of medical students to image-guided surgery and surgical simulation.

Soft segmentation allows each pixel to belong to multiple classes, with varying degrees of membership. More common segmentation methods use hard segmentation where a pixel is assigned exclusively to one class. Hard decisions about the membership of a pixel involve throwing away some of the information contained in the data; soft segmentation methods have the advantage of avoiding these decisions until later in the process, thereby keeping more options available for post-processing steps. Soft segmentation can be converted to hard segmentation by using the maximum membership classification rule. Under this rule a pixel is assigned to the class with which it has the highest membership value.

In this work we segmented the data in  3 classes and pixels were categorized as belonging to low intensity bone, brain matter, and cerebrospinal fluid (CSF). Four soft segmentation algorithms were applied and their results compared.

The input scan consists of 3D CT images of size 512 x 512. In the volumetric data, the voxel size is 0.48mm x 0.48mm x 4.5mm (width x height x depth). To visualize our results, we show the output from soft segmentation as RGB images, the corresponding hard segmentation as grayscale images, and we generate 3D models of the brain from the resultant hard segmentation.

CT Brain Data

Brain segmentation of CT images. Red for low intensity bone, green for brain matter, blue for cerebrospinal fluid and ventricles. (a) Original axial CT slice. (b) Bayesian classification. (c) Fuzzy-c Means. (d) Population Diameter Independent algorithm. (e) Expectation Maximization.