Since CT provides detailed, cross-sectional views of all types of tissue, it is one of the best tools for examining the chest and abdomen. It is also the preferred method for diagnosing many different cancers. CT can clearly show even very small bones as well as surrounding tissues such as muscle and blood vessels. This makes it invaluable in diagnosing and treating spinal problems and injuries to the hands, feet and other skeletal structures. Another most known CT usage is diagnosis of cerebrovascular accidents and intracranial hemorrhage also know as “Head CT” or “CT Brain”. CT can also play a significant role in the detection, diagnosis and treatment of vascular diseases that can lead to stroke, kidney failure or even death.
Major project goals are to:
The general assumption of our experiments was that general structure of healthy human anatomy is common. The main idea was compare “sick” CT scan image with generalized “healthy” CT scan image combined from finite amount of CT scan of “healthy” organs (e.g. Brain CT), and show results on the input image by marking the “pathology” regions. The algorithm had two basic step:
Both of the examined averaging techniques provided extremely bad results,while median was a little bit better. Edge detection was found to be too accurate and discovered lots of areas, but none pathology ones. The reason is that both images(the healthy and the damaged one) were not taken from the exactly same angle and the same person.The distance transform algorithm was one level better than edge detection. Since brain structure is very complicated, edge detector (the preprocessing stage for distance transform algorithm) found too many edges. Hence each pixel of the pathology was not far enough in order to pass desired threshold, or if the threshold was not high enough, we recieved the exact database image. Segmentation was entirely different approach since we had no need in database image. The segmentation method via relaxation labeling provided relatively good result. Pathology areas of specific kind were found. Although,segmentaion method could not ignore the bones of the skull and declared them as pathology. Still, despite the fact the results were ambiguous, part of the pathologies were discovered.
We have to make several important observations from the previous assays in this art of field: segmentation methods were used at the precomputation stage in each one of the approaches. By combining this observation with our results, we conclude that the segmentation is a very basic tool for computer-aided diagnosis(CAD),in particular, CT. We have surveyed the studies that handled CT scans of the following organs:brain,liver,lungs and abdominal. Dispite the fact that every one of them used the segmentation method as the basic tool at the early stage, rather distinctive directions were taken afterwards. And that is for the following reasons: