人内耳结构的三维精确建模研究/Computer-aided Three-dimensional Precise Modeling of Human Inner
内耳是听觉和平衡觉的感受器。其包含的壶腹嵴、囊斑和膜迷路等结构的三维形态和复杂的空间关系是其功能的重要基础。采取单纯切片和常规解剖的方法，难以准确地对其进行形态学的研究。内耳的听觉和平衡觉的感受功能依赖于基底膜的振动，内淋巴液流动、耳石膜的运动等生物物理过程。由于内耳深藏于颞骨骨质中、结构及其复杂和细微，使这些生物物理过程难以观察和记录。计算机技术的发展为内耳的形态和生理学研究提供了新的手段。利用火棉胶切片可以对内耳的结构进行三维重建研究。虚拟现实技术的发展也使内耳生物物理研究出现了新的思路。但以往的三维重建研究受到计算机硬件、软件性能的限制，几乎都是选取部分切片用于内耳部分结构的形态学研究。本实验的目的是建立包含壶腹嵴、囊斑等内耳细微结构的精确的内耳三维模型，以期用于内耳生物物理学的虚拟现实研究和内耳细微结构的三维形态学研究。本实验采用新鲜人颞骨，经固定、脱钙、脱水和火棉胶包埋，滑动切片机连续切片，获得了片厚20um和25um的火棉胶连续切片两套。获取内耳结构二维图像采用了三种方法。第一种是不染色的切片经幻灯机投影放大后，用高像素数码相机拍摄整体图像。第二种是每张投影机放大后图像，用数码相机分四部分拍摄，然后用CANON UTILITIES PhotoStitch 3.1软件进行图像拼合。第三种是用与解剖镜耦联的数码相机直接连续拍摄火棉胶包埋块的切面。分别获得内耳结构的连续二维图像3套。结果显示：不染色的火棉胶切片图像可以分辨出需要重建的囊斑、壶腹嵴等结构；图像拼合后的图像更加清晰；第三种方法获得图像较前两种方法略差，也可以分辨以上精细结构；采用第三方法时，酒精硬化的火棉胶包埋块的切面图像比香柏油处理的清晰。本实验使用第二种方法获得全套图像，在高性能个人电脑上，利用Able software 3D-DOCTOR Version 3.5软件进行三维重建，其切片厚度为20um，经过设定图像标度参数设定、配准、分割等过程，建立了包含壶腹嵴、囊斑等精细结构的精确的内耳的整体三维模型。不同结构用不同颜色显示，每一结构通过不透明、透明、网格化等不同方式进行显示，可以任意选择不同的结构显示。获得的内耳结构的各种三维模型使用VRML建模语言输出，在CORTONA VRML浏览器中实现了桌面虚拟现实演示。本实验证明：使用不染色的切片可以进行内耳结构的精确三维建模，却可以避免染色过程对切片的损耗和引起的切片变形，以致影响模型的精确性；采用高像素数码相机使获得二维图像更加清晰；3D DOCTOR软件可以在个人电脑上进行复杂的内耳结构的三维建模，同时重建多个结构，同时或分别进行显示，但其自动分割功能由于内耳结构的复杂无法应用，分割仍然需要人工进行。重建的三维模型显示：椭圆囊斑主要部分与水平半规管在同一平面上，球囊斑的主要部分与上半规管在同一平面上。重建的三维模型用VRML建模语言输出后可以在普通个人电脑上实现虚拟现实演示，可以用于教学和形态学研究。本实验建立的精确的内耳结构三维模型将进一步用于内耳生物物理过程的虚拟现实研究。
Inner ear is the sensory organ of hearing and equilibrium. Its three-dimensional configuration and conformation, which is very important to its function, is extremely complex. It is difficult to perform precise morphologic study by normal sections and traditional dissection. The sensory function of hearing and equilibrium is based on some biophysical processes, for example the vibration of basilar membrane, the flow of endolymphatic fluid and the motion of otolithic membrane. Inner ear is deeply located in the temporal bone, at the same time its structure is extremely tiny and complex, which makes these processes is very hard to observe and record. The development of computer science provides the morphologic and physiological study of inner ear some new methods. Three-dimensional reconstruction with serial celloidin sections could be applied to the morphologic research of inner ear. The technology of Virtual Reality produces new ideas about biophysical research of inner ear. But with the limitation of the capability of hardware and software, former three-dimensional reconstructions of human inner ear usually base on part sections, for example every 5 to 10 sections. And those studies often focus on the morphology of local inner ear structures. The objective of this study is to establish precise inner ear models containing tiny structures, for example the crista ampullaris, macula utriculi and macula sacculi. Then perform the virtual Reality research of inner ear biophysics and morphologic research of inner ear by these models. Four fresh human cadaveric temporal bones were harvested, fixed, decalcification, dehydration and embedded by celloidin. After sectioning by sliding microtome, two sets of serial inner ear celloidin sections without staining were acquired, with thickness of 20 um and 25 um. Three methods were used to acquire the two-dimensional digital image of inner ear. The first one was capturing the whole unstained image projected and magnified by projector with Digital Camera. The second one was dividing the whole image into four parts to capture, and then merging them into one clearer whole image by the software of CANON UTILITIES PhotoStitch 3.1. The third one was capturing the cutting plane of celloidin block by DC connected to dissection microscopy. Eventually, three sets of serial two-dimensional digital images of inner ear were acquired. The results show that vestibular end organs were easy to differentiate in celloidin sections without staining. The images merged by four part images are much clearer than those captured wholly. With the third method, the images were not as clear as those acquired by other methods, but tiny structures were also able to differentiate. When third method was used, the images from the celloidin block hardening by alcohol are clearer than those from block hardening by cedarwood oil. The total serial images acquired by the second method were imported to the software of 3D-DOCTOR Version 3.5, which cutting thickness is 20um. The process of three-dimensional reconstruction was performed on a personal computer with high capability. In the 3D-DOCTOR software, the images were calibrated, aligned and segmented. Then a whole inner ear model containing the crista ampullaris and macula was produced. Different structures were displayed by different colors. Every structure could be displayed by different methods, fox example opacity, transparency and wire frame, et al. Any structure could be elected to display or not. The 3d model was exported by VRML (Virtual Reality Modeling Language). With the CORTONA VRML viewer, the Desktop VR displaying of the model could be achieved on personal computer. Some conclusions could be made by this research. With unstained celloidin sections, precisely 3d modeling of inner ear could be accomplished. 3d model from unstained celloidin sections might be more precise because the destruction and distortion of sections were avoided. By the DC with high pixels, the 2d images of inner ear are much clear. By the software of 3D-DOCTOR, complex 3d reconstruction work of inner ear could be performed on personal computer. At one time, more than 10 structures could be segmented, reconstructed and displayed. But the function of auto-segment was not smart enough. The main work of segment was still done manually. From the 3d model of inner ear, it was found that the main part of macula utriculi and the horizontal semicular canal were on the same plane, while the main part of macula sacculi and superior semicular canal were on the same plane. After export by VRML and displayed in VRML viewer on personal computer, the Virtual Reality display of 3d inner ear model could be achieved, which benefited education and morphological research. These precise 3d models would be further applied to VR research of inner ear biophysics.