Method for automatic tissue segmentation of medical images
A method to segment images that contain multiple objects in a nested structure including acquiring an image; defining the multiple objects by layers, each layer corresponding to one region, where a region contains an innermost object and all the objects nested within the innermost object; stacking the layers in an order of the nested structure of the multiple objects, the stack of layers having at least a top layer and a bottom layer; extending each layer with padded nodes; connecting the top layer to a sink and the bottom layer to a source, wherein each intermediate layer between the top layer and the bottom layer are connected only to the adjacent layer by undirected links; and measuring a boundary length for each layer.
1. A method of segmenting objects in a nested structure comprising:
initializing an amplitude distribution of each object to obtain estimated distributions;
segmenting an image using a nested graph cut (NGC) with estimated distributions to obtain new segmentation results;
iterating between segmentation using NGC with said estimated amplitude distributions and distribution estimation based on said new segmentation results; and
continuing said iteration until said estimated amplitude distribution or segmentation converge; and
when an erroneous segment part of an image segment exists, a post processing step to remove the erroneous segment part of the image segment, wherein the post processing step comprises:
selecting a highest volume component of a segmented part;
using the highest volume component to determine an acceptable volume percentage;
selecting a next highest volume component of the segmented part and comparing the volume to the acceptable volume percentage, wherein if the volume is higher than the acceptable volume percentage, the next highest volume component is connected with the segmented part; and
continuing comparing the volume components of different segment parts until the next highest volume component of the different segment parts is lower than the acceptable volume percentage.
2. The method as recited in claim 1 wherein the segmentation is based on intensity information.
3. The method as recited in claim 1 wherein a final round of segmenting an image uses intensity distributions obtained from full resolution segmentation.
4. The method as recited in claim 1 further comprising a spline based curve fitting process to estimate a depth dependent mean and variance caused by acoustic attenuation and focusing.
5. The method as recited in claim 1 wherein iterating between segmentation using NGC includes using a fitted means and variance profile and identifying inlier and outlier variations, wherein only the inlier variations are used.
6. The method as recited in claim 5 wherein the inlier variations of the fitted means and variance profile are used in a next segmentation iteration.