Jagoe J R, Paton K A
Br J Ind Med. 1975 Nov;32(4):267-72. doi: 10.1136/oem.32.4.267.
Computer programs for measuring simple pneumoconiosis in radiographs are described and assessed. The 36 films studied had been read by 11 skilled human observers and a 'radiological score' of pneumoconiotic severity was therefore available for each film. The computer assigns to each square grid of side 3-6 mm a measure which reflects the unevenness of the density distribution in that grid. The 'computed score' is defined as the mean diversity over all relevant grids in both lung fields. On the set of 36 films the correlation between radiological score and computed score was 0-88. By contrast, the correlation between the score assigned by a single observer and the average of the scores assigned by the other 10 was in the range 0-95 to 0-98. The program can use the computed score to classify a film into one of the four major International Labour Office (ILO) U/C categories, the success rate of this process being 80% compared with those quoted by other workers in the range 45%-65%. If the films used in this study be typical, then the program described may form the basis of an automatic method for measuring pneumoconiosis in epidemiological work.
本文描述并评估了用于在X光片中测量单纯尘肺的计算机程序。所研究的36张胸片已由11名技术熟练的专业人员阅片,因此每一张胸片都有一个尘肺严重程度的“放射学评分”。计算机为每个边长为3 - 6毫米的方形网格赋予一个数值,该数值反映了该网格内密度分布的不均匀性。“计算评分”定义为双肺野所有相关网格的平均差异度。在这36张胸片中,放射学评分与计算评分之间的相关性为0.88。相比之下,单个观察者给出的评分与其他10位观察者评分均值之间的相关性在0.95至0.98之间。该程序可利用计算评分将胸片归类到国际劳工组织(ILO)的四个主要U/C类别之一,此过程的成功率为80%,而其他研究人员报道的成功率在45% - 65%之间。如果本研究中使用的胸片具有代表性,那么所描述的程序可能成为流行病学工作中测量尘肺的自动方法的基础。