Research On The Method To Extract Characteristic Parameters Of A Fetal Head Based On Ultrasound Image Technology

In order to evaluate fetal intrauterine growth, ultrasound imaging technique is generally adopted in the clinic as an important inspective and diagnostic method for obstetrics. BPD and OFD is one pair of the most important parameters that are applied for evaluating fetal intrauterine growth. The values of BPD&OFD, which are obtained with ultrasound imaging system, are used for understanding the situation of fetal weight and physical development in obstetrical practice.BPD and OFD are measured through sonographer'manual operation, there are always big errors of the manual measurements. The accuracy of evaluating fetal intrauterine growth depends heavily on the sonographer'experience. As a result, it is meaningful to introduce s techniques of semi-automatic or automatic feature exaction into ultrasound image processing, so as to improve the accuracy of parametric measurement and the diagnostic effect, and meanwhile to reduce the medical work intensity.In order to extract feature from ultrasound images semi-automatically or automatically, requirements of high SNR and clear features of target ultrasound imagine must be satisfied. This dissertation focuses on two crucial problems in the ultrasound image-image denoising and image segmentation, aims at the semi-automated or automated image analysis and computer-aided diagnosis in the obstetric ultrasound image. The studies have been carried out in the following aspects:(1) Image denoising. Based on the survey and comparative research on the current image denoising techniques, we focus on the speckle reducing anisotropic diffusion(SRAD) technique. Because of SRAD's low convergence and limited denoising ability in the homogeneous part of image caused by only meeting some and not all the requirements of designing SRAD's diffusion expression, a new diffusion expression is proposed, which not only meets all the requirements of designing SRAD's diffusion expression but also obtains satisfying experimental results.(2) Image segmentation. Basd on review of technologies on image segmentation, Fuzzy c-means(FCM) is studied in detail. It is later found that diffusion thresholds of improved FCM algorithms are too accurate and make the results of edge detection are sensitive to the tiny edge,so that it will go ill with the work of extracting primary edge. As a result, a traditional FCM model to extract fetal head's contour is adopted in this dissertation and satisfying experimental results have been gotten. The fetal head's contour can be extracted semi-automatically in the ultrasound image.(3) According to the fact that current fetal head's feature extraction system take no consideration to the fetal head's deformation effect produced by squeezing and other reasons, a GVF Active Model is used to extract fetal head's contour after the FCM's processing on the image. In order to alleviate GVF Active Model's sensitivity to the initializing position and to quicken operational speed, the energy function of GVF Active Model has been modified. The modified GVF Active Model provides an edifying attempt on the task of extract fetal head's contour. However, the research of applying Active Model on medical ultrasound image is still in exploratory stage and there are many problems to be solved before applying it in clinic.