AUTOMATIC SEGMENTATION AND CLASSIFICATION OF CELLS FROM BRONCHO ALVEOLAR LAVAGE

Olivier Lezoray, Michel Lecluse

Abstract

Broncho alveolar lavage is the most commonly used diagnostic tool for confirming alveolar hemorrhage. Golde has introduced a ranking score, based on the hemosiderin content of macrophages which enables ranking cells from 0 to 4 based on the degree of Prussian blue stain. We propose a complete image analysis scheme to automatically perform both the extraction of the cellular objects and the ranking of each cell according to the Golde score. The image analysis techniques used mainly involve clustering and mathematical morphology. A 2D histogram is clustered to extract the main cellular components, a color watershed is used to determine and refine the regions. Finally, the cellular components of interest are firstly classified according to their hue and secondly according to their staining repartition. The proposed image analysis technique is very fast and produces reliable and accurate results.

Keywords
clustering; color; Golde score; mathematical morphology; segmentation

Full Text:

PDF


DOI: 10.5566/ias.v26.p111-119

Image Analysis & Stereology
EISSN 1854-5165 (Electronic version)
ISSN 1580-3139 (Printed version)