Segmentation of object of interest from medical images has been a challenging task in the field of medical image analysis to identify diseases. The medical images can be acquired from Magnetic Resonance Images (MRI), Computer Tomography (CT) or X-ray scanners. The raw images inherently contain noise, generated during the acquisition process. As such, image noise has to be filtered at the pre-processing stage. Moreover, the filtered image may contain tissues belonging to several classes. Therefore, the noise filtered image has to be classified to a certain number of classes or regions which is determined heuristically.

The classified image contains the object of interest as well as other structures that need to be removed. The primary focus of this research is to remove such unwanted structures from the classified image and extract only the object of interest. For this purpose, it is proposed to utilize deformable models which exhibit profound flexibility during the segmentation process. After the initial curve is initialized at the classified image, such model can be made to evolve around the object of interest while tracking image information such as intensity levels, edge strength, etc.

This research group mainly focuses on how to extract anatomical structures of interest from any medical image.