Standardized & Fast Rendering For Medical Image

At present, the medical image diagnosis of our country is gradually entering into a period of comprehensive digitalization. Various kinds of digital imaging equipments get assembled in hospitals. PACS (Picture Archiving and Communication System) is also growing in popularity, and the regional image comprehensive platform appears in some developed areas. This means huge amounts of digital image will be displayed on different equipments by different practitioners in different regions, which makes the following problems become more and more obvious:1) The same digital image may not get consistent display results caused by characteristics of different display device and the difference of element such as surroundings. This brings more challenge to image diagnostic, and more inconvenient to communication of image reading;2) The amount of image data from a single image check is becoming larger and larger, and doctors have a higher demand in fast and accurate image rendering.3) The new image application technology emerge consistently, especially the diagnosis and surgical navigation based on 3-D topographic images, which make the real-time image rendering more and more critical.This paper, starting from the above problems, spreads out the research on fast standardizing rendering of medical images. The main contents are organized as the following:1) Solve the problems of standard gray-scale images consistency display through Grayscale Standard Display Function (GSDF).At present, most medical imaging diagnostic system cannot satisfy the consistency display of DICOM (Digital Imaging and Communications in medicine).The outstanding performance is unable to perform a linear relationship between the P-Values and eye-visual perception. We proposed the following procedures to improve the visual perception effect of medical images:This paper proposed a fast GSDF algorithm based on gray-scale display function of DICOM PS3.14.First,acquire the mapping relationship between the DDL(Digital Driving Level) value and brightness value, Then, get the corresponding relationship between 3-spling interpolation performance value and DDL value, according to Barten model, Finally, obtain the display effect achieved by storage value, with the help of adjustable window operations.The algorithm can rapidly realize standard gray-scale display function, and make great good linear relationship between image performance values and eye-visual perception, acquire DV/BV between great detail variance and background variance. Use this method can compensate most commercial display devices'insufficient of DICOM P14 correction, and can also calibrate and quality inspect to medical devices which have their own P14 calibration function, which has significant application value.2) Realize the fast standardization draw of DICOM gray-scale images, according to acceleration technology based on graphics hardware, with the integration of GSPS (Grayscale softcopy presentation sate)and GSDF,This paper aim to realize fast standardized rendering of DICOM gray-scale images by integrating the GSPS and GSDF. The method is:Produce texture data by modality values, which is the output by stored values through device lookup table, load the texture data to graphics hardware; If the VOI lookup table is configured during display process, make the VOI table into 1-D texture, then load it to the graphics hardware; Otherwise, convert the window adjust operation into the lookup table,and the load it; Make the performance value into 1-D texture then load it; Calibrate the standard gray-scale display function, convert it into lookup table, then make 1-D texture and load it into graphics hardware; Acquire the ultimate digital drive level values through corresponding lookup table according to the convert process of graph coloring stage based on the technology of graphic hardware acceleration, finish painting,then completes the image rendering.Using the above methods, with the help of hardware acceleration technology, we can greatly improve the real-time GSPS operation, and have apparently advantage compared to the traditional adjustable window technology. What's more, This method is not confined to the current application of 256 levels DDL, and it can adjust GSDF lookup table according to the hardware equipment, thus it break the limitation of operating system to gray image which play best utility of standard display device.3) Realize the texture-mapping-volume-rendering algorithm with good portability using GLSL(OpenGL Shader Language),and solve the efficient problem under common graphics hardware a high quality volume rendering using classic lighting model.Through the improving conventional texture-mapping-volume-rendering algorithm, we achieved good light effects of 3-D display.And on this basis, conventional texture mapping volume rendering the limit physical space is proposed based on GLSL (OpenGL shading language),which can be in common PC general graphics hardware medical data field on a high quality volume rendering technique of effective methods. Use programmable fragment shader of GPGPU, the texture data quantity will be a quarter of traditional methods including a one-dimensional transfer function texture and an equivalent to the source data scale 3d texture, effectively save the physical memory.Through the contrast experiment analysis, the method can be obtained in ordinary graphics hardware of good real-time and higher quality volume rendering effects.4) Solve the problems of fast rendering of mass data with common graphics hardware based on the partitioned strategy of Full Binary Tree.First, adopt full binary tree to realize block texture mechanism, and carry on the scientific mathematical description, reduce texture to update the frequency of Secondly, use a inheritance relationship of abstract block strategy, deduce the mathematical methods of solving general, lower boundary the complexity of the calculation; Finally, make the transfer function to be one-dimensional texture using GLSL,and adopt the data field size equates to body work dynamic texture and set the shading stage through fragment shades search and realize dynamic lighting calculation, again reduce texture memory possessive amount, effectively improve the large-scale data field volume rendering real-time.Experimental results show that the dynamic texture working set using abstract block and inheriting relationship management boundary, through calculation of the ways to reduce the hierarchical structure of partitioned drawing calculation and dynamic texture control complexity. Using this algorithm can be in ordinary PC for far more than the texture memory of large-scale body data complete have good real-time and higher quality volume rendering.5) Make the above algorithm coded into library according 4 levels using cross-platform of OpenGL through object encapsulation, and realize the software system integration and application on PC system.After strict theoretical arguments and experimental tests, the technology and the algorithm has been applied to the author unit software products PACS system and platform Vamos 3D COMPLEX, with good practical value.