按照Mandelbrot的分形理论,医学图象及许多自然图象的灰度表面的形成均符合分形布朗运动规律,而 且可以用分形的维数来表征图象灰度表面的精细与粗糙程度。文中正是基于这种思想,采用图象的分形维数作为一个特征参量,对人体的肌肉组织进行超声定征。对60多个样本三类病变图象提取分形维数,并采用基于Bayes法则的分类器分类,实验表明:用分形维数对组织进行定征,正确率达88.33%。这为医学的临床辅助诊断提供了一种新的参考量,对提高病变诊断的正确率有重大的意义。

Mandelbrot' s fractal theory regards that many natural and medical images as the end result of Fractional Brownian Motion(FBM) . The fractal dimension is suit to measure the fineness or roughness of the surface.Based on this idea, the fractal dimension is applied in classifing ultrasonic images of muscle organ in this article. After the fractal dimensions of 60 sam- ples are calculated by this way ,then classified by Bayes classifier,the correct classification rate accounts to 88.33%。The very high rate provethat the fractal dimension can be a new feature for clinical diagnose and very significant to improve the correct diagnostic rate.