Optimization Method Of Transfer Function For Volume Rendering In Medical Images

Scientific visualization is the discipline of automatically rendering images from scientific data. Adequate visual abstractions are important to show relevant information in the data. Medical visualization have important meaning in the medical science study. The volume rendering techniques are the important point of Scientific visualization. Direct volume rendering can present all information of the data as image. Direct volume rendering because of the advantages of imaging results have been widely used and research.In volume rendering, transfer function (TF) map samples of 3-dimensional data to optical parameters.TF directly determine the effect of rendering and become the key of volume rendering. However, there are two major problems of TF's design: because of lacking of an intuitive user interface, user often require a lot of attempts of defining data fields, repeatedly adjust visualization parameters and spend a lot of time and energy for interest objects with TF; The design of TF is blind as lacking of data field's guidance information. Sometimes even a good TF has been designed to get the best rendering results, but you may not know. In the past decade, there is still not a good way to solve this problem. To some extent, the difficult of finding a suitable TF hinder the wider application of volume rendering. Looking for a good TF has been identified as visualization problems in one of the top ten. Research seems quite important and urgent on the method of TF's design.Currently, the first target of TF's research is developing an intuitive user interface so that adjustment of transfer function parameters is more convenient and the design is more efficiency; The second target is providing meaningful instructions to assist users of designing good TF for reducing the blindness of the design; The third target is the automatic design of appropriate TF for different data fields which making TF's design trend of automation and intelligent. Now, there are some studies that focus on the goal for improving efficiency of volume rendering. The researchers lead the technology of artificial intelligence into TF's design process to improving the efficiency of volume rendering. While the intelligent technology similar with people's smart, so leading intelligence technology into volume rendering process also makes the results meeting customer's need and improves the usefulness of volume rendering. In recent years it has gradually become a new research topic of large-scale scientific data visualization. It is important for the development of visualization.In this article, we discuss the transfer function for volume rendering and research the method to optimize it with genetic algorithm. The main work of this paper are summarized as follows:(1) The design method of transfer function based on IGAThis paper focuses on the TF's design based on traditional genetic algorithm (GA).The main section includes TF encoding/decoding, the generation of middle TF and the fitness calculation of TF corresponding images. There are some problems of GA,like a weak local search ability and convergence early and so on, resulting in the final result from the TF's design process is not ideal. But the local search ability and stability of the fast local adjusting genetic algorithm(FLAGA) are more strong, so this paper proposes an improved genetic algorithm (IGA) which combine the advantages of GA and FLAGA. Main improvements have been made in the individual selection, genetic crossover and mutation in order to have the GA's global search ability and FLAGA's local search ability. The calculation is executed with the steps of the design method of TF based on GA, using image boundary entropy as Image quality evaluation standard, and applying the IGA algorithm into design of TF. We analysis the performance of the GA, FLAGA, IGA optimization algorithms by experiment and verify the validity of the IGA algorithm in TF's design.(2) The design method of transfer function fusing multiple featuresBased on image-center method, some scholars proposed a design framework for integrated the TFs which allows users to choose interest characteristics directly in a number of images with volume rendering displaying a single structural features. The framkword can automatically adjust the TF's parameters and retain the number of selected features. Finally it get a fused image which display a variety of organizational structures, while retaining the interesting features. The core idea of this framework is change the problem of integration of multiple TF into the energy function minimization problem based on the similarity of edge images. It uses the traditional genetic algorithm to solve the optimization problem. But there is some problems, like slow convergence, premature convergence and bad interactivity and etc.In order to better solve the rapid integration of a variety of organizational characteristics, this paper draws on the advantages of the design framework and propose a new integration of the transfer function of optimal design framework. We explore the TF fusion method of keep many interesting features and optimize the design process. The new framework greatly reduce the computational complexity and accelerate processing speed. Main features are:the IGA algorithm is introduced to the framework for enhancing the search algorithm ability and accelerating the search speed; to design a new energy function expressions and a new method of the similarity calculation of edge images for more accurate evaluation of fusion Image; introduction of the evaluatior and setting various end conditions for allowing users to choose to stop or automatic stop and achieving better interactivity.Our framework compose by of the TF generator, the volume rendering device, the energy function calculator and the evaluatior. At first we choose the source transfer function and send to TF generator. The TF generator is initialized according to the source transfer function, using IGA algorithm to adjust the TF, and send the results to the volume rendering device for volume rendering. Then the intermediate rendered images which generated by the volume rendering device are sent to energy function calculator for calculating the image similarity and get the corresponding energy value. The energy function calculator send the energy value to the evaluator. The evaluator assess the effect has been satisfied. Energy function is used to represent the fitness function of genetic algorithm. The smaller energy value represents a more suitable intermediate transfer function. Based on these energy values, the TF generator eliminate inappropriate TF and use IGA algorithm to generate a new set of TF. And then the framework starts the next cycle of assessment. When the evaluator getting the satisfactory result, the best TF which system output shows the best volume rendering effect.(3) Integration and application in 3-D processing systemMy laboratory design a set of three-dimensional medical image processing system, called Vamos 3D COMPLEX, in order to meet daily needs. The major purpose of the system is to solve the problem of new algorithms integration. When new researchers contact with medical imaging related algorithms, they often need to learn lots of background knowledge and receive much training, such as coding and other related skills. These mentioned above undoubtedly increase the working time that can be avoided. So we develop the system to provide a platform for algorithm integration with the common interface to minimize work capacity.The main features is the use of plug-in design, so researchers can easily use function modules to acqure necessary 3-D display and related processing, enabling them to focus their eyes on the algorithms.The full functionality of the system covers the complete two-dimensional image processing, reconstruction, multiplanar reconstruction and virtual endoscopy and etc. The system has good scalability and interactivity to support DICOM standard, the three-dimensional mouse and positioning equipment.The main work of my participation are:the needs analysis of volume rendering module, including the detailed analysis of parameter tuning, configuration management, algorithms management, cutting function and protocol management function; the design of database tables, involving to study records, image storage, thumbnail, etc.; data interface module and data processing module, responsible for reading all kinds of image data and format conversion; agreement management module, in accordance with the study part to manage the configuration of algorithm and related parameters used in volume rendering; the design and implementation of volume rendering viewer widget, using MVC (Model-View-Controller) model to display images of three-dimensional reconstruction, support to display and control image annotation and call protocol and other functions; multiplanar reconstruction capabilities, selected any sense of line or curve Regions of interest to obtain the corresponding surface reconstruction image; the design and implementation of the module of TF's fusion design, integrating into Vamos system and achieve the function of fusing a variety of organizational characteristics.