Design And Implementation Of Gallbladder Segmentation System In Abdominal CT Images

Gallbladder segmentation in CT images is usually needed for gallbladder disease diagnosis and surgical planning work.In the case of automatic methods can not achieving desired results,the design and development of the gallbladder interactive segmentation system is necessary.This system can help doctors quickly and accurately complete the split task.Based on the analysis of the requirement of gallbladder segmentation,a gallbladder segmentation system was designed.The functions of the system include image data reading,image data visualization in three views which are axial plane,sagittal plane and coronal plane,window width and center adjustment,interactive seed selection and parameter settings,regional growth based on neighborhood statistics,interactive growth of target surface contiguous,interactive processing of segmentation results,and so on.In view of the CT value of ligaments and bile ducts in the gallbladder and its periphery,the problem of segmentation leakage is very easy to occur by using the threshold-based growth method.A regional growth method based on neighbor-hood statistical information is proposed.A voxel can become a target point,only when sufficient voxels in the neighbor of this voxel have CT values in the threshold range,thus limiting the growth to the small organization.In order to solve insufficient segmentation problem of above method,two kinds of contiguous growth methods-continuous growth with fixed threshold and continuous growth with adaptive threshold,are proposed.A voxel on the surface of the segmented target has potential to grow if its CT value is in the range of low and high threshold.Those potential voxels are connected together to grow if the number of voxels is enough.The fixed threshold method is faster,but the accuracy is lower than that of the adaptive threshold.Two ways have their own advantages and disadvantages,the user can choose according to requirement.The system provides an interactive processing tool such as a flat cut,an eraser,a fill brush,etc.,allowing the user to patch the split result.The results show that the regional growth and target surface growth method based on neighborhood statistical information can be used to obtain good segmentation results.After a small amount of interactive treatment,satisfying result can be achieved.