Discriminating bacterial spores from inert airborne particles by classification of optical scattering patterns

Scattering patterns are made available by the TAOS (Two-dimensional Angle-resolved Optical Scattering) method, which consists of detecting micrometer-sized single airborne aerosol particles and collecting the intensity of the light they scatter from a pulsed, monochromatic laser beam. TAOS patterns have been classified by a learning machine, the training stage of which depends on many control parameters. Patterns due to single bacterial spores (Bq class) have to be discriminated from those produced by outdoor aerosol particles (Kq set) and diesel soot aggregates (sq set), where both Kq and sq are assumed not to contain patterns of bacterial origin. This work describes two directions along which classification continues to develop: the enlargement of the control parameter set and the simultaneous processing of two areas (sectors) selected from the TAOS pattern. The latter algorithm is meant to make the classifier sensitive to simmetry exhibited by some patterns. The available classification scheme is summarized, as well as the rule by which discrimination is rated off-line. Discrimination based on one pattern sector alone scores fewer than 15% false negatives (misclassified Bq patterns) and false positives from Kq and sq. Discrimination based on the symmetry of two pattern sectors fails to recognize 30% of the Bq (bacterial) patterns, whereas < 5% Kq (environmental) patterns are assigned to the Bq class; false positives from sq (diesel) patterns drop to zero. The issue of false positives is briefly discussed in relation to the fraction of airborne bacteria found in aerosols. © 2014 SPIE.