Research Of Intracranial Facial Nerve Image Segmentation Algorithm Technology For Three-Dimensional Reconstruction

Hemifacial Spasm, is a middle-aged brain nerve diseases, accompanied during the onset of facial muscles twitch and pain, especially in everyday conversation and smile more obvious, so the disease can seriously affect the patient’s daily life. The current treatment of the disease is locating the relationship between responsibility vessels and facial nerve through analysis of MR sequence image by doctors, and then using micro-vascular decompression surgery (MVD) for clinical treatment of the lesion area. Since the cause of the analysis relies mainly on the doctor’s subjective sense analysis, resulting in a lack of visual analysis of the etiology model. This topic is in this background presents a three-dimensional visualization of the facial nerve brainstem surface adjacent to blood vessels, and this article mainly for the research of facial nerve segmentation.By analyzing the characteristics of MR image sequences which is provided by cooperation unit, this article will be divided into four steps for achieving the segmentation of facial nerve:(1) Image interpolation. As the facial nerve tissue in the brain appear as thin shape, and MR image sequences is0.5mm to scan, it will lead to fewer sheets for facial nerve in MR image sequences, which is lacking of three-dimensional spatial information, is not conducive to the subsequent three-dimensional reconstruction, so first of all to the MR image sequence interpolation between slices and reconstruction of the missing three-dimensional information.(2) Image enhancement. Due to the lower resolution of the MR image sequences and the noise in imaging process, the facial nerve area is difficult to be distinguished, so we need to enhance the image, which will improve the contrast between the facial nerve area and surrounding issue area and improve the accuracy of segmentation.(3) Seeds iterative region growing. Because it is the segmentation of MR image sequence, each image is an independent division will definitely lead to inefficient algorithms. Using the continuous relevance of facial nerve in the three-dimensional space, this article proposed the idea of the seed point iteration, which is combined with the traditional region growing algorithm and is effectively segment the facial nerve area.(4) The processed images. After the above segmentation is only preliminary results, so the image must be processed to remove interference and noise. This article presents the three-dimensional region growing algorithm, which using the facial nerve of three-dimensional spatial information, as a post-processing algorithms.Proved by experiments, this article presents the algorithm which is according to facial nerve characteristics in MR image sequences is effective and can accurately segment the facial nerve area. This algorithm is more simple and fast than the traditional segmentation based on MR image sequences and establishes the foundation for the subsequent reconstruction of three-dimensional models.The innovation of this article:(1) Using the continuous relevance in MR image sequences, the facial nerve segmentation algorithms merge the segment operation and rebuilt operation together simultaneously. The dividing provide the direction and the rebuilding provide the termination conditions. The segmentation algorithm can determine more effective termination conditions and can better avoid over-segmentation.(2) Using the three-dimensional spatial information in MR image sequences, the post-processing algorithm for preliminary segmentation can constrain the noise in three-dimensional space and can more effectively remove the interference from the preliminary segmentation.