Research On The Preprocessing And Segmentation Of DICOM Image Sequence Based On CUDA

With the development of imaging technology for medical inspection, medical image iswidely applied to the clinical diagnosis and adjuvant therapy which could help medicalworkers to explore the histological structure of organs directly. DICOM standard is developedto unify the discrepancy of storage format and transmission protocol of medical imagesgenerated by difference medical equipment. It has been developed into an internationalstandard of image and relevant information transmission between devices in medical imaginginformatics. The number of the sequences of DICOM images which include large numbers ofdata is increase with the medical devices. How to process the extremely large amounts ofDICOM image data efficiently and provide the user-friendly tools for medical workers toextract the regions of interest is becoming the investigative emphasis and hot spot of medicalimage. In this paper, the pretreatment and segmentation algorithm of medical image which isreliable and efficient has been studied to reduce the workload of medical workers and makethem more productive.The multi-scale display of the DICOM medical image is the important step of detectingthe lesions and making the radiotherapy plan. It is achieved by the dynamic regulation of thescale which also called window operation to observe the organ tissues with different densities.The traditional way utilizing the linear function fails to be consistent with the characteristicsof the human eye and loses image details. In this paper, two kinds of nonlinear multi-scaledisplay method are utilized to extend the gray level of the region of interest. In order to solvethe contradiction between the long computing time of the nonlinear function and the real timeof the dynamic window operation, the parallel generation algorithm of the image isimplemented on the CUDA platform. The experiment results show that the optimizationmethod is at least an order of magnitude faster than classical algorithms and achieves thepurpose of image enhancement which provides fundamental base for the further research ofthe DICOM medical image.The generation of medical image is easily influence by time and space. The medicalimage contains much noise and uncertainty. The traditional hard image segmentation methodis difficult to achieve the desired result. The fuzzy classification technique can processing theuncertainty in medical image. But it contains large computational quantity and can’t guaranteethe real time. The method of histogram and the technology of general purpose computing ongraphics processing units are utilized in this paper. The better initial cluster centers are givento reduce the iterations. We realize the parallelization of the image segmentation algorithm onthe GPU platform to reduce the computational time and ensure the real time. The experimentresults show that the method segment the medical image with prospective effect. By the works mentioned above, the system of the preprocessing and segmentation ofDICOM image sequence is completed which provide the interactive functional for healthworkers to view the medical image. The whole system could show the region of interest in theDICOM image efficiently and intuitively in multiple. The visual quality has been improved soas to achieve the better segmentation result. The system is user-friendly and easy to use whichhave high practical value.