We are working with medical CT image datasets. The purpose of this job is to develop an automated function that can find and label all the bone structures within a volume. The program will take a dicom file list or a 16bit volume data as the input and the label mask will be provided as the output. Simple threshold techniques do not take bone interior into account, so morphological or similar postprocessing functions are required to label porous regions. Moreover, angiographic images are problematic when using threshold techniques since blood vessles have similar gray levels with bones in such images. The technique should also be able to separate such vessels and label the bones only while omitting the vessels. In case of labelling inaccuracies,it should be possible to provide seedpoints to the algorithm that indicate incorrectly labelled regions. In such cases the program should reiterate the labelling and produce new results. The total processing time for a 512x512x512 volume should not exceed 10 seconds using an i7 8 core CPU. The process progress should be provided with a callback function.
You are not required to develop a UI. We only need the code for segmentation developed in C++ preferrably using VTK (not a must).The sample datasets can be provided upon request.