Feature Extraction And Quality Analysis Of Liver Ultrasonic Images

Ultrasound image diagnosis is an important medical image diagnosis means like x-ray CT, isotopr scanning and MRI. The diagnosis about fatty liver on the base of ultrasound image is the important methord. Hower , compared with other mdeical images(e. g. CT and MRI), the quality of ultrasound images is relatively low.At present,the diagnosis about fatty liver is determine the nature, the conclusion of disease is influenced by doctors .In this paper , both feature extraction and classification recognition are mainly studied , on the base of extracting statistical parameter , we apply fractal geometry theory to receive fractal parameter , and combined to the fatty liver image's character gain dendity of light dots by image segmentation .On the base of successful gainning textrue feature of ularsonic, we designded BP nerve network . We select a great deal of swatch to train nerve network , and gain the best right value. Using this nerve network ,we classify the fatty liver from normal liver and obtrain good effect.