Detection of pollen grains in multifocal optical microscopy images of air samples

Abstract

Pollen is a major cause of allergy and monitoring pollen in the air is relevant for diagnostic purposes, development of pollen forecasts, and for biomedical and biological researches. Since counting airborne pollen is a time-consuming task and requires specialized personnel, an automated pollen counting system is desirable. In this article, we present a method for detecting pollen in multifocal optical microscopy images of air samples collected by a Burkard pollen sampler, as a first step in an automated pollen counting procedure. Both color and shape information was used to discriminate pollen grains from other airborne material in the images, such as fungal spores and dirt. A training set of 44 images from successive focal planes (stacks) was used to train the system in recognizing pollen color and for optimization. The performance of the system has been evaluated using a separate set of 17 image stacks containing 65 pollen grains, of which 86% was detected. The obtained precision of 61% can still be increased in the next step of classifying the different pollen in such a counting system. These results show that the detection of pollen is feasible in images from a pollen sampler collecting ambient air. This first step in automated pollen detection may form a reliable basis for an automated pollen counting system.

Features extraction techniques for pollen grain classification

“…The next 10 systems correspond to the application of each transformed domain over the geometrical and texture parameters using again both as classifiers. In the third approach, 10 more systems were implemented corresponding to a series of fusion systems, which decision was based on the Adding-Score (AS) technique applied at the decision level [15]. The supervised learning strategy was implemented in all the experiments.…”

“…An interesting fully automatic pollen grain detector and classifier was presented in [15]. The system used a multifocal microscope to obtain the pollen grain images from air samples.…”

Pollen segmentation and feature evaluation for automatic classification in bright-field microscopy

“…images from ambient air, a previous image cleaning from dirt, fungal spores and other non-pollen particles (Landsmeer et al, 2009) is required. This is also a time consuming process where a robust automatic segmentation is a challenging problem.…”

Automated Spore Measurements Using Microscopy, Image Analysis, and Peak Recognition of Near-Monodisperse Aerosols

“…The morphological criteria assumed that releases of interest would consist of uniformly sized, nearly spherical particles. Bacterial and fungal spores in the respirable size range tend to be spherical or nearspherical, and researchers developing an automated method to detect pollen grains in ambient air samples also assumed a circular shape for these particles in optical microscopy images (Landsmeer et al 2009). …”

“…Another refinement would be the inclusion of other particle features, e.g., color and internal structures, to automatically identify detected spores. Multifocal image-processing and pattern recognition techniques have been used to identify microbial colonies (Puchkov 2010) and pollen grains (Chen et al 2006;Ranzato et al 2007;Landsmeer et al 2009). Recent advancements that allow the efficient acquisition of large volumes of image data also should be explored, such as rapid digital cameras, automated motorized microscopes, and increased computational power.…”