Research On Intracranial Hematoma Image Segmentation

Intracerebral hemorrhage is one of the diseases with the highest mortality and disability rate in the world.Surgery is an effective treatment method.Doctors need to segment the image of intracerebral hematoma before making the operation plan,usually using manual segmentation method,but manual segmentation is time-consuming and laborious.Therefore,this thesis improves the fast marching(FM)method,designs a three-dimensional full-automatic CT image segmentation method of intracerebral hematoma,and names it iFM method is used to assist doctors to make operation plan.The iFM method is improved in the following two aspects.Firstly,based on threshold segmentation,the iFM method designs a method to automatically obtain seed points from volume data.It solves the problem that the FM method needs to set seeds manually,and improves the FM method from semi-automatic segmentation to full-automatic segmentation.Secondly,the iFM method improves the segmentation efficiency while ensuring the segmentation accuracy.The iFM method reduces the amount of data by calculating the bounding box of three-dimensional region of interest,and then builds the pyramid of three-dimensional region of interest,iteratively segmenting from the top of the pyramid to the bottom of the pyramid,so as to speed up the segmentation of the algorithm.In addition,three-dimensional visualization of segmentation results and estimation of hematoma volume can help doctors to diagnose.In order to verify the effectiveness of the iFM method,this thesis compares the method of automatically obtaining seed points with other methods of automatically selecting seed points.At the same time,taking the results of expert manual segmentation as the gold standard and four widely used evaluation indexes as the basis,the segmentation results of the iFM method are quantitatively compared with the two newer active contour model segmentation methods in recent three years.The experimental results show that the accuracy of the iFM method to automatically obtain seed points is higher.In addition,the iFM method also has excellent performance in segmentation accuracy and speed.It not only has the highest accuracy in one of the four evaluation indexes,but also has the highest segmentation speed.The research features and innovations of this thesis are as follows:(1)The method of building the pyramid of three-dimensional region of interest is proposed,which is combined with FM method to realize the fast iterative segmentation in the pyramid and reduce the amount of data and the times of evolutionary computation.The experimental results show that the iFM method has a significant improvement in the efficiency of segmentation compared with FM method.(2)A method of automatically obtaining seed points from volume data is proposed.Compared with other methods of automatically selecting seed points,this method is less affected by noise,more seed points are selected,and then the segmentation efficiency of FM method is improved.Experimental results show that this method is more accurate than other methods of automatically selecting seed points.