Assessment of bronchial inflammation using an automated cell recognition system based on colour analysis.

The aim of this study was to develop an automated system of cell recognition based upon colour analysis suitable for microscopic examination of bronchial inflammation. Human bronchi obtained from 17 patients undergoing thoracotomy were embedded in glycolmethacrylate to perform immunohistochemistry with antibodies against: neutrophil elastase, tryptase, chymase, eosinophil cationic protein, CD68, CD3 and immunoglobulin E. The image analysis system calculates three independent criteria (optic density, hue density, hue) combined with morphological parameters to specifically recognize a positive staining. This automated analysis was applied to the study of bronchial inflammation in smokers and nonsmokers in terms of the absolute number of cells and the expression of different markers by a single cell. The use of these criteria enabled the characterization of a positive stain on single (intraclass correlation coefficient (ICC) = 0.88 or serial (ICC = 0.84) sections. Cell counts obtained by the automated system were highly reproducible. Regarding bronchial inflammation, it was found that the number of inflammatory cells was significantly higher in smokers than in nonsmokers, the majority of these cells bearing immunoglobulin E. These results demonstrate that such computerized analysis of colours is a valuable method for quantifying inflammatory cells in bronchial tissue and for analysing the expression of different markers by a single cell.