Mishima M, Oku Y, Kawakami K, Sakai N, Fukui M, Hirai T, Chin K, Ohi M, Nishimura K, Itoh H, Tanemura M, Kuno K
Department of Clinical Physiology, Kyoto University, Japan.
Front Med Biol Eng. 1997;8(1):19-34.
An automated and quantitative assessment of the spatial distribution of low attenuation areas (LAA) on X-ray CT was performed using texture analysis in chronic pulmonary emphysema (CPE) patients. LAA was defined as those areas having a density less than the mean minus one SD of the control (-960 HU). The probability of change from non-LAA to LAA between a given pair of pixels with horizontal intervals of i pixels (RNi) was evaluated, because this reflects the interaction between LAA and non-LAA regions with different resolutions. The relationship between the percentage area of the LAA over the total area of the entire lung field (LAA%) was subsequently estimated. The RNi increased sharply as the i value increased from 1 to 5, but then almost became a plateau for i values larger than 5. This suggests that the fundamental structures in the LAA areas ranged from 1 x 1 to 5 x 5 pixels in size. RN1-LAA% and RN5-LAA% plots produced curves which were convex, with peak values at approximately 50 LAA% of 0.09 and 0.18, respectively. In the RN5/RN1-LAA% plot, the RN5/RN1 ratio remained constant at 2.0 regardless of the LAA%. A random process simulation was performed to determine the patterns of LAA proliferation if the spatial distribution of the LAA units was random. When the unit size was kept constant, the results of the simulation did not fit the empirical relationship between the LAA% and the three parameters (RN1, RN5 and RN5/RN1). The simulation provided the best-fitting curves when the unit size of the LAA increased in proportion with the LAA%, starting from a 1 x 1 pixel size increasing at a ratio 1 x 1/(5 LAA%). This suggested that the LAA units do not proliferate randomly in spatial orientation at a fixed unit size, but rather spread throughout the whole lung field in a congregated form whilst increasing their unit size. Thus, it may be concluded that healthy lung tissues near emphysematous lesions have a high probability of suffering from emphysema in the future. This may be due to a direct effect of the neighboring emphysematous lesion or due to a pathologic change in the larger bronchii which dominate both the healthy tissues and the emphysematous lesions.
在慢性肺气肿（CPE）患者中，利用纹理分析对X线CT上低衰减区域（LAA）的空间分布进行了自动化定量评估。LAA被定义为密度低于对照组平均值减一个标准差（-960 HU）的区域。评估了水平间隔为i个像素的给定像素对之间从非LAA转变为LAA的概率（RNi），因为这反映了不同分辨率下LAA与非LAA区域之间的相互作用。随后估计了LAA面积占整个肺野总面积的百分比（LAA%）之间的关系。随着i值从1增加到5，RNi急剧增加，但当i值大于5时，RNi几乎趋于平稳。这表明LAA区域的基本结构大小范围为1×1到5×5像素。RN1-LAA%和RN5-LAA%图产生的曲线呈凸形，峰值分别出现在约50 LAA%时的0.09和0.18。在RN5/RN1-LAA%图中，无论LAA%如何，RN5/RN1比值均保持在2.0不变。进行了随机过程模拟，以确定如果LAA单元的空间分布是随机的，LAA增殖的模式。当单元大小保持恒定时，模拟结果不符合LAA%与三个参数（RN1、RN5和RN5/RN1）之间的经验关系。当LAA的单元大小从1×1像素大小开始以1×1/（5 LAA%）的比例增加，并与LAA%成比例增加时，模拟提供了最佳拟合曲线。这表明LAA单元在固定单元大小下不会在空间取向上随机增殖，而是以聚集的形式在整个肺野中扩散，同时增加其单元大小。因此，可以得出结论，肺气肿病变附近的健康肺组织未来患肺气肿的可能性很高。这可能是由于邻近肺气肿病变的直接影响，或者是由于支配健康组织和肺气肿病变的较大支气管的病理变化。
