Fromson B H, Denison D M
Lung Function and Computed Tomography Unit, Brompton Hospital, London.
Thorax. 1988 Feb;43(2):120-6. doi: 10.1136/thx.43.2.120.
This study set out to determine whether quantitative features of lung computed tomography scans could be identified that would lead to a tightly defined normal range for use in assessing patients. Fourteen normal subjects with apparently healthy lungs were studied. A technique was developed for rapid and automatic extraction of lung field data from the computed tomography scans. The Hounsfield unit histograms were constructed and, when normalised for predicted lung volumes, shown to be consistent in shape for all the subjects. A three dimensional presentation of the data in the form of a "net plot" was devised, and from this a logarithmic relationship between the area of each lung slice and its mean density was derived (r = 0.9, n = 545, p less than 0.0001). The residual density, calculated as the difference between measured density and density predicted from the relationship with area, was shown to be normally distributed with a mean of 0 and a standard deviation of 25 Hounsfield units (chi 2 test: p less than 0.05). A presentation combining this residual density with the net plot is described.
本研究旨在确定能否识别出肺部计算机断层扫描的定量特征，从而得出一个严格定义的正常范围，用于评估患者。对14名肺部明显健康的正常受试者进行了研究。开发了一种从计算机断层扫描中快速自动提取肺野数据的技术。构建了亨斯菲尔德单位直方图，在对预测肺容积进行归一化后，所有受试者的直方图形状显示一致。设计了一种以“网状图”形式呈现数据的三维展示方式，并从中得出每个肺切片面积与其平均密度之间的对数关系（r = 0.9，n = 545，p < 0.0001）。计算得出的残余密度为测量密度与根据面积关系预测的密度之间的差值，结果显示其呈正态分布，均值为0，标准差为25亨斯菲尔德单位（卡方检验：p < 0.05）。描述了一种将此残余密度与网状图相结合的展示方式。
