Gradient correction and classification of CT lung images for the automatedquantification of mosaic attenuation pattern

Purpose: The detection of density differences, or "mosaic attenuation patte rn," on CT images may be difficult when the regional inhomogeneity of the d ensity of the lung parenchyma is subtle. The purpose of this work was to de velop a fully automated method for the reproducible quantification of the u nderattenuated areas of the lung parenchyma. This technique may be useful i n increasing the precision of investigation of structure/function relations hips. Method: Anatomical segmentation was achieved by a structure-filtering opera tor based on mathematical morphology. To compensate for the density gradien t visible on lung CT scans, a model-based iterative deconvolution filter an d an adaptive clustering algorithm were developed. Validation was performed with CT images from a lung phantom, 15 patients with constrictive oblitera tive bronchiolitis, and 8 normal subjects. Results: The accuracy of the estimate of the density gradient on phantom st udies was 93.3%. The automated quantification of the areas of decreased att enuation on scans of constrictive obliterative bronchiolitis was within 8.2 % from the average scoring of two experienced observers. Conclusion: The proposed technique is fully automated and can accurately co rrect for density gradient and classify areas of decreased attenuation on l ung CT images.