One of the oldest life sciences applications for vision system technology is medical imaging. To help researchers, physicians and surgeons identify specific information on an image, medical image segmentation is used. Image segmentation involves dividing the CT scans, MRIs, and X-ray images, for example, into multiple parts to identify relevant information that is then used to diagnose, provide therapy, and perform surgery.
Image Segmentation Methods
Otsu’s method is used to preprocess images before analysis. Nonuniform background illumination in an image is corrected and then the image is converted into a binary image. This allows for easy identification of foreground objects. It is then possible to count the objects, find the area of the objects, and compute statistics.
Pathologists stain tissues with hematoxylin and eosin. This staining helps them to distinguish between different tissue types. This results in a white, blue, and pink tricolor image. This image is then processed to separate out the colors into three separate images. Now a pathologist can view the development of only a single type of tissue.
Watershed segmentation is used to separate touching objects in an image. The image is first converted to grayscale. Then the foreground and background objects are marked or color-coded to help identify separate objects.
Medical Image Segmentation Applications
Brain Tumor Image Segmentation
Image segmentation of a brain tumor is challenging. The variability of target structures in the brain limits the use of a priori knowledge. Human supervision of 3D segmentation of the brain tumor helps to estimate the volume and spread of the tumor. A probability map is created to localize the tumor. Then active contours are drawn to delineate the tumor.
Prostate Image Segmentation
Semi-automatic prostate segmentation helps physicians recommend an approach to treat prostate cancer. A robotic system able to deal with image noise uses ultrasound imaging to find the borders of the prostate, enabling a reliable calculation of prostate volume. This information is essential to planning surgeries, radiation therapy, cryotherapy, and high-intensity focus ultrasound.
Skeletal Image Segmentation
Only cortical bone has the density needed to differentiate it from soft tissues on an X-ray. Porous spongy bone offers only low X-ray attenuation and is equal to or only slightly higher than soft tissues.
To segment the image, a histogram is built to differentiate between bone and soft tissue. This is done with an automatic segmentation procedure where a binary value of 1 is given to bone and 0 to soft tissue. The procedure is repeated so that at every iteration the value of the intensity of the threshold is increased.
Kidney Image Segmentation
Computed Tomography (CT) scans help surgeons plan operations and decrease surgery risks and duration. A technique has been designed to create a kidney model for each patient. Four CT scans are completed, each with a different contrast level.
The first image has no contrast. The following three scans are enhanced each with an increasing level of contrast agent. Combining these images provides a clear CT scan of the kidney arteries, veins, and ureter. A 3D model of the kidney is built out of the images. Physicians are able to simulate surgery alternatives and compare their feasibility and possible results.
Advances in vision system technology are expanding medical imaging capabilities. Keep up to date by visiting VisionOnline.