基于超声图像特征区分子宫肌瘤和腺肌病

超声图像子宫肌瘤和腺肌病的区分目前主要依赖于医生的经验,缺乏客观的指标.为提高区分的性能,提出一种新的对子宫肌瘤和腺肌病的超声图像进行多分辨率分析的自动分类方法.提取图像在多分辨率下的纹理参数,同时结合计算出的带方向分形维数,建立支撑矢量机进行子宫肌瘤和腺肌病的分类判决.通过对27例正常、45例腺肌病和74例肌瘤离体样本超声图像进行分析,结果表明：提取的多分辨率纹理参数和带方向的分形维数对区分子宫肌瘤和腺肌病是敏感的,结合这两类参数建立的支撑矢量机区分子宫肌瘤和腺肌病的正确率近100%.

The Classification of Uterine Myoma and Uterine Adenomyosis Based on Ultrasound Image Features

The classification of the uterine myoma and the uterine adenomyosis from ultrasound images mainly depends on doctors' experience and lacks objective criterions by now.A novel automatic classification method is proposed to improve the performance.The multiresolution analysis was done for ultrasound images of the uterine myoma and the uterine adenomyosis to obtain their texture parameters under various resolutions.Together with the orientational fractal dimension parameters,a Support Vector Machine(SVM) was established to classify the uterine myoma and the uterine adenomyosis.The result of the experiments,in which there were 27 normal cases,45 adenomyosis cases and 74 myoma cases,showed that multiresolution texture parameters and orientational fractal parameters were both sensitive to the uterine myoma and the uterine adenomyosis.The classification accuracy of the myoma and the adenomyosis based on SVM with all these parameters is about 100%.