Neural networks for increased accuracy of allergenic pollen monitoring

Abstract

Monitoring of airborne pollen concentrations provides an important source of information for the globally increasing number of hay fever patients. Airborne pollen is traditionally counted under the microscope, but with the latest developments in image recognition methods, automating this process has become feasible. A challenge that persists, however, is that many pollen grains cannot be distinguished beyond the genus or family level using a microscope. Here, we assess the use of Convolutional Neural Networks (CNNs) to increase taxonomic accuracy for airborne pollen. As a case study we use the nettle family (Urticaceae), which contains two main genera (Urtica and Parietaria) common in European landscapes which pollen cannot be separated by trained specialists. While pollen from Urtica species has very low allergenic relevance, pollen from several species of Parietaria is severely allergenic. We collect pollen from both fresh as well as from herbarium specimens and use these without the often used acetolysis step to train the CNN model. The models show that unacetolyzed Urticaceae pollen grains can be distinguished with > 98% accuracy. We then apply our model on before unseen Urticaceae pollen collected from aerobiological samples and show that the genera can be confidently distinguished, despite the more challenging input images that are often overlain by debris. Our method can also be applied to other pollen families in the future and will thus help to make allergenic pollen monitoring more specific.

The key role of contact time in elucidating the mechanisms of enhanced decontamination by Fe0/MnO2/sand systems

“…Iron corrosion determined the extent of MB discoloration and the intuitive increase of MB discoloration with increasing Fe 0 loading is observed in all systems only after a pseudo-steady state is established in the systems or the capacity of MnO 2 is exhausted. Alyoussef 59 , 70 , 71 tested a parallel system with 4.5 g L –1 of MnO 2 and observed a larger decrease of MB discoloration for 18 days. Similar observations were made by Noubactep et al 82 in their experiments for uranium removal.…”

“…This is achieved herein by investigating the systems for 7, 18, 25 and 47 days. This corresponds to following the fate of aqueous MB (discoloration) as the contact time increases from 7 to 47 days 59 , 70 , 71 . In particular the variation of the pH value in the systems will be discussed in detail.…”

“…Hybrid systems tested as means to prevent iron passivation include amendment with granular activated carbon (GAC), magnetite (Fe 3 O 4 ), manganese oxides (MnO x ), pyrite (FeS 2 ), and sand 15 . Among these aggregates, MnO x and FeS 2 are the most chemically reactive 36 , 71 . Both aggregates induce a pH shift to more acidic values.…”

“…Due to the good adsorptive affinities of Fe 2+ and Fe 3+ for sand surface, it is assumed that the deposition of Fe hydroxides at its surface (coating) will compete with Fe 2+ consumption by the reductive dissolution of MnO x until sand coating is completed. Thereafter, the “free” precipitation of iron and manganese occurs and the final products are deposits of more or less pure iron and manganese ores 66 , 71 . In other words, the investigated Fe 0 /MnO x /sand system is a ternary system only at the start of the experiment.…”

Development and application of a method to classify airborne pollen taxa concentration using light scattering data

“…Automated methods for evaluating airborne pollen concentrations have been developed by various research groups 1 . Examples include image recognition by deep learning 2 – 4 , neural networks 5 , fluorescence spectrometry 6 , microscopy image classification 7 , 8 , and nuclear magnetic resonance spectra 9 . Specifically, Plair Rapid-E is an automated pollen recognition device that uses scattering information and fluorescence spectra 10 .…”