| Medical ultrasound is the preferred technique due to its irreplaceable advantages: safe,valid, portable and cost effective. However, the image quality of ultrasound is low with itsimaging mechanism. Therefore, how to obtain high quality ultrasound images, accuratecontours of the region of interest and quantified son graphic features for the radiologists todiagnose the deseases more objectively have been urgent to be solved. This paper mainlyresearch ultrasound image segmentation and features selection. Detail works are listed asbelow:1. Feature extraction methods of Ultrasound images are studied. Two main properties ofthe image pixel are gray value and texture value. Two approaches to extract texture feature ofmedical ultrasound image were used in this paper which are GLCM based method and theLaws of the energy filter templates. Gray value and the extracted texture value of pixel areextracted to be the input of the pulse coupled neural network.2. Pulse coupled neural network is used to extract the initial contour of ultrasound image.Using pulse coupled neural network (PCNN) in image processing and analysis can achievebetter results by simulating the visual characteristics of mammalian. However, mathematicalmodel of PCNN has many parameters such as threshold factor, the time decay constant, thelink factor and so on. These parameters control the operation efficiency of the network. A newmethod for parameter selection is proposed in this paper. Then the improved pulse coupledneural network is used for the initial ultrasound image contour extraction.3. The improved active contour model for accurate segmentation of ultrasound images isproposed in this paper. Traditional active contour model can not segment the image withcurrent complex structure effectively. A differential evolution algorithm based on cloud modelis used to improve active contour model. The experimental results show that the improvedalgorithm can effectively segment the lymph node ultrasound images with complex structure. |
PDF Full Text Request