Differential Equations Seminar

Dr. Giovanni Franco Crosta from the University Milan-Bicocca

Location

Mathematics/Psychology : 401

Date & Time

May 12, 2014, 11:00 am12:00 pm

Description

Title: Can one discriminate bacterial spores from inert airborne particulate matter ?


Abstract: The discrimination of aerosolized biological threat, represented by bacterial spores, from harmless, respirable particles is a basic need in environmental monitoring. A by now well-known experimental method, TAOS (Two-dimension Angle-resolved Optical Scattering), detects single, airborne material particles, which are then illuminated by a monochromatic laser pulse. The corresponding scattering pattern covers a wide solid angle, although the phase and polarisation of the scattered wave are ignored. TAOS patterns from microspheres of known materials, from single bacterial spores, soot aggregates and from outdoor aerosol particles have been collected and stored by the thousands. The recognition of TAOS patterns by application of inverse scattering theory remains a challenge. In 2013, algorithms based on machine intelligence have discriminated bacterial spores from all other materials within the limits of 20% false negatives and less than 11% false positives. More recently, discrimination has improved by addressing two issues: 1) the adjustment of a detection threshold, which controls the interfering (a.k.a. ”confounding”) role played by diesel soot aggregates; 2) the simultaneous processing of two regions of interest in the TAOS pattern in order to enhance pattern symmetry, if any. One implementation of the classifier has yielded false positives and false negatives in the 15% range. In another implementation, at the expense of more false negatives, interference from confounders has been brought to zero.