Renal Artery Segmentation Based On Three-Dimensional Zernike Moments

With the increase of kidney cancer incidence year by year, the operation of renal tumors gets more and more attentions. To know the spatial relationship of renal artery and renal tumors is a key step for renal tumor resection operation. So accurate segmentation of renal artery is very important. The kidney CTA (Computed Tomographic Angiography) image has huge amount of data, manual segmentations produce lots of works for doctors. In order to improve the efficiency and diagnosis accuracy, a semi-automated segmentation method based on 3D Zernike moments was proposed for renal artery segmentation in CTA images.Firstly, three-dimensional vessel segmentation methods are summarized. The initial vessels can be efficiently extracted by the threshold region growing algorithm, but there are under and over segmentations. In the CTA image, the local arterial vascular shows a bright 3D tubular structure, but background structures are different. The local geometric structure is extracted by threshold region growing method, and 3D Zernike moments are used to form 3DZD (Three-Dimensional Zernike Descriptors) feature vector for the local structure feature. The vessel segmentation classifier was trained by Support Vector Machine classification algorithms. In order to improve the processing efficiency, the region of interest was extracted by the global fixed threshold region growing algorithm and the extension regions of the main vessel branch centerlines. The amount of the data was further reduced. The segmentation results were classified from the region of interest, which were connected with the main vessel branches to extract the complete blood vessel.The proposed method was verified by 30 CTA images. The measure was defined by the similarity of the segmentation results and the experts reference centerlines in evaluation framework. In the analysis part, global and manual threshold region growing methods were compared with the proposed method. The results show that the proposed method has higher accuracy. The proposed method can be used to extract renal artery from CTA images and provides basic conditions for clinical application.