分形维数特征量对病变组织的超声定征

Classification of the Sick Organise Based on the Feature of Fractal Dim ension

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。

Mandelbrot's fractaltheory regardsthatm any naturaland m edicalim agesas the end resultofFractionalBrow nian Motion(FBM) . Thefractaldim ension issuitto m easure the finenessorroughnessofthe surface.Based on thisidea, the frac taldim ension isapplied in classifing ultrasonicim agesofm uscle organ in this article. Afterthe fractaldim ensions of60 sam ples are calculated by this way ,then classified by Bayesclassifier,the correctclassification rate accountsto 88.33% 。The very high rate provethatthe fractaldim ension can be a new feature forclinicaldiagnose and very significantto im prove the correct diagnosticrate.