Knowledge Guided Automatic Contour Initialization And Segmentation Of Abdominal Structures In CT Images

• Main Author: Jawarneh, Mahmoud Saleh Mahmoud
• Format: Thesis
• Language: English
• Published: 2012
• Subjects: QA1 Mathematics (General)
• Online Access: http://eprints.usm.my/45658/

Computed Tomography (CT) scans are becoming a priceless means of diagnosing abdominal structures. CT scans result in a huge number of 2D slices of the acquired anatomical part in abdominal imaging. CT are more preferred compared to sensitive imaging techniques such as MRI in abdominal imaging owing to their high signal to noise and good spatial resolution. In the area of medical image processing, the current interests are in the automated analysis and visualization of liver, spleen, and kidney to assist in diagnosis, radiation therapy planning and surgical planning. Delineation of these structures which is still an open research problem is the first and fundamental step in all of these studies. Automation of medical image segmentation reduces time-consuming, tedious, subjective human interaction tasks and may aid radiologists, who are normally required to view thousands of images daily. Thus, automatic segmentation is the main focus of several research efforts. In this research, we propose an automatic knowledge-based segmentation framework based on active contour methods. The proposed segmentation system is generic, and employs multiple sources of medical knowledge: medical atlas; expert’s rules; multiple views: axial, coronal and sagittal; image features and image DICOM Meta data. The focus in this research is on level set active contour segmentation methods which provide promising results, robust to dataset variations and do not require extensive prior training. As such, they can be reliably used for segmentation of major structures in abdominal CT scans. The obtained results are very promising showing significant improvements over other methods where the volume measurements error is 7% and the processing time was improved by 68%.