Diagnostic imaging of acute pulmonary embolism.

Abstract

The common strategy of combining clinical information, lung scintigraphy and pulmonary angiography in the diagnosis of acute pulmonary embolism (PE), has many limitations in clinical use. The major causes are that pulmonary angiography and lung scintigraphy are not universally available, and that pulmonary angiography is very expensive. The purpose of this thesis was to analyse different aspects of validity in regard to lung scintigraphy, pulmonary angiography, spiral CT, and ultrasound of the legs, with the subsequent intention of discussing new diagnostic strategies. Observer variations in lung scintigraphy interpretation when applying the PIOPED criteria were tested in 2 studies with 2 and 3 observers respectively and expressed as kappa values. The ability to improve agreement in lung scintigraphy interpretation was tested by training 2 observers from different hospitals. The impact of 3 observers' variations in lung scintigraphy interpretation when compared to pulmonary angiography, was tested by comparing the ROC areas of the observers. The value of combining subjectively derived numerical probabilities and the PIOPED categorical probabilities in lung scintigraphy reporting was compared to using the PIOPED categorization only, and this was tested by comparing ROC areas. The sensitivity and specificity of detecting an embolic source in the deep veins of the legs by ultrasound as a sign of PE when lung scintigraphy is inconclusive, was tested by comparison with pulmonary angiography. The sensitivity and specificity of spiral CT, compared to pulmonary angiography, was tested by comparison to pulmonary angiography. The inter- and intra-observer kappa values were in the range of moderate and fair. It was not possible to achieve better kappa values after training. Although observer variations were substantial, the accuracy did not differ significantly between the 3 observers. Incoorporating subjectively derived probabilities into lung scan reporting could not reduce the number of inconclusive investigations. Sensitivity and specificity of ultrasound in detecting PE was 0.70 and 0.97, respectively. However, 2 patients (of 9) had deep venous thrombosis and no pulmonary emboli at angiography. The sensitivity and specificity of spiral CT was 0.90 and 0.96, respectively. The observer variations at lung scintigraphy are substantial and may be difficult to improve between hospitals, even though the accuracy of observers in general is good. Although subjectively derived interpretation criteria did not show to be useful when added to categorical interpretation criteria, they may be useful when substituting established criteria. Despite recent progress in refining interpretation criteria, a substantial fraction of the patients still need pulmonary angiography to be performed. However, in many patients pulmonary angiography is not performed as prescribed. Spiral CT and ultrasound of the legs is a new favourable diagnostic strategy with a high validity in detecting venous thromboembolic disease, and a good availability and cost-effectiveness.