Automatic Whole Heart Segmentation And 3D Interactive Display Software Based On CTA Images

The aortic vessels that surround the heart are called as coronary artery,which can cause Coronary Heart Disease(CHD)when it is narrow or even blocked.Coronary Computed Tomography Angiography(CTA)is needed to diagnose if the patients have CHD.The whole heart can be mapped into a series of two-dimensional images by CTA for doctors to diagnose CHD.However,there are some challenges for diagnosing CHD with CTA images,for example,interferences can be found in CTA images,heart can hardly be presented in threedimensional through two-dimensional CTA images,and so on.In view of the advantages of computer aided image processing,it is necessary to utilize computer for segmenting and modeling CTA images to improve the accuracy and reliability for diagnosing CHD.There are some methods and software at present used to assist the diagnosis of CHD by processing CTA images.However,some problems arise while using these software to improve doctors’ efficiency for there is no one-click segmentation scheme and lack of customization for doctors’ demands in these software.Aiming at the above challenges in these CTA image processing software,in this thesis,we propose a one-click scheme for whole-heart segmentation.Based on this scheme,a software platform is developed based on CTA images for fully automatic whole-heart segmentation and 3D interactive presentation.The main contributions of this paper are as follows:(1)An improved regional growth segmentation algorithm is proposed in this thesis based on CTA images to achieve automatic segmentation of whole heart by only one click.We solve the problems of connectivity to get a better segmentation result by limiting gradient consistency and regional similarity,making multiple areas grow at the same time and detecting narrow points.The empirical value,grey level histogram and cross-correlation method are utilized in this scheme to segment the whole heart automatically,also some preprocessing and post-processing methods are used.The segmentation results based on this scheme achieve an accuracy of 0.85 Dice scores and 0.7 Intersection of Union(Io U).(2)A software platform for automatically displaying and segmenting CTA images has been designed and developed in this thesis.Most functions of this software platform can be completed with one click,avoiding the complicated process of parameters adjustment by users.3D interactive display,axes rotation and distance measurement are all implemented in this platform for better helping doctors diagnose CHD.The software platform developed in this thesis has been implemented and deployed in the imaging and other relevant departments of the hospital.After nearly four months of practical use by clinicians,the various functions of the platform have been verified.The platform successfully solves the problems doctors faced in operation and achieve great clinical results.