Computer Vision News - May 2016

Tool The SimpleITK Toolbox COMPUTER VISION NEWS 12 Insight Segmentation and Registration Toolkit (ITK) is an open-source, cross- platform system that provides a substantial set of software tools for image analysis. ITK employs state of the art algorithms for segmentation and registration and it includes a built-in integration for medical imaging data providing a support for multidimensional imaging data. The goal of SimpleITK is to provide an abstraction layer to ITK that enables developers and users to access the powerful features of the Insight Toolkit in a more simplified manner. It provides a simple layer to ITK's complicated C++ templated API to be easily wrappable in several languages, to name a few: Python, Java, C#, Lua. SimpleITK was created in 2010/2011 as part of a concerted effort to simplify the use of ITK and make it more accessible to a wider audience. In particular, it aims at satisfying the needs of a community interested in rapid prototyping, education and use of image analysis from interpreted language. During the last year, the SimpleITK has gained stability and included a wider range of features: in particular, Version 0.9 was released in October together with a SimpleITK workshop held at the last MICCAI conference in Munich, Germany. To get a taste of the things you can do with SimpleITK, we will review two classic examples of medical image processing: (a) how to perform lung segmentation from a CT image; (b) registration of two spinal vertebrae. The demo will be in Python, and we'll run them interactively with ipython from the browser. The codes here are partially based on the tutorial SimpleITK workshop. Details about the installation can be found here . In the first example we will examine Region Growing , the simplest (or one of the simplest) image segmentation technique. Region growing segmentation iteratively examines neighboring pixels from initial pre-selected seed points and determines (based on some similarity definition) whether the pixel neighbors should be added to the region. The definition of pixels similarity is what varies between the different forms of the region growing technique. Let’s start with the code - Read data and select seed point(s) First thing first, (after successful installation) you have to import the SimpleITK model along with the ‘gui’ model which we will also need for this denominations. The code snippet for this is below: Next, you will need to load some data to work with. In our example we will perform a lung segmentation so we will load an example of CT scan (you can find such files here ) .