Quantitative assessment of radiation-induced lung changes by computerized optical densitometry of routine chest x-rays

Purpose: To develop a quantitative assay of radiological lung changes after postmastectomy radiotherapy applicable for analysis of routine chest x-rays. Methods and Materials: The assay relies on measurement of the optical density of chest x-rays by means of a standard personal computer-controlled film densitometer used for scanning dosimetry films. Density profiles were recorded at 1 mm steps along two reference lines in each lung. The crossing between the clavicula and the first rib in the x-ray projection was used as an easily identifiable anatomical landmark and was used to establish two parallel cranio-caudal reference lines separated by 20 mm. The starting point for recording optical densities was 5 mm below a line perpendicular to the reference lines and tangent to the top of the lung. Data were stored in a computer file for subsequent processing. The optical film densities in the apex of the lungs were converted into equivalent absorber thickness (EAT). To minimize the dependency on technical factors, the unirradiated lung was used as a control. Lung density changes were quantified by the relative change in EAT (REAT), which was evaluated as the difference between the summed EATs along the reference Lines in the two lungs, corrected for any pretreatment difference in lung density, and taken as a percentage of the pretreatment EAT value of the lung. Results: Four different tests were carried out to investigate the reproducibility and validity of the proposed assay. (a) An anthropomorphic phantom[;was used to test the relationship between REAT and the layer of plastic absorber placed in front of one lung. A linear relationship was found with a correlation coefficient of 0.9993. (b) A series of 10 repeat measurements of the same arbitrarily chosen x-ray showed a mean REAT value of 9.8%, with a standard deviation of 0.21%. (c) Forty-three chest x-rays exposed on the same day were available from the clinical series. These were treated as double determinations of REAT values, and the standard deviation was estimated at 1.35%. (d) REAT values estimated with this assay were significantly correlated with the scores of two experienced specialists, an oncologist and a radiologist (Spearman's rank correlation coefficient being 0.605, p < 10(-8)). Conclusion: This assay quantifies radiation-induced lung injury from routine chest x-rays. Thus, it is possible to take advantage of the very large retrospective materials available in most oncological institutions. The assay has been validated in a phantom experiment and shown to be reproducible. Furthermore, it is technically easy to perform.