Publication Date:
2024-01-12
Description:
Monitoring of airborne pollen concentrations provides an important source of information for the
\nglobally increasing number of hay fever patients. Airborne pollen is traditionally counted under
\nthe microscope, but with the latest developments in image recognition methods, automating this
\nprocess has become feasible. A challenge that persists, however, is that many pollen grains cannot
\nbe distinguished beyond the genus or family level using a microscope. Here, we assess the use of
\nConvolutional Neural Networks (CNNs) to increase taxonomic accuracy for airborne pollen. As a case
\nstudy we use the nettle family (Urticaceae), which contains two main genera (
\nUrtica
\n and
\nParietaria
\n)
\ncommon in European landscapes which pollen cannot be separated by trained specialists. While
\npollen from
\nUrtica
\n species has very low allergenic relevance, pollen from several species of
\nParietaria
\nis severely allergenic. We collect pollen from both fresh as well as from herbarium specimens and
\nuse these without the often used acetolysis step to train the CNN model. The models show that
\nunacetolyzed Urticaceae pollen grains can be distinguished with 〉
\n98% accuracy. We then apply our
\nmodel on before unseen Urticaceae pollen collected from aerobiological samples and show that the
\ngenera can be confidently distinguished, despite the more challenging input images that are often
\noverlain by debris. Our method can also be applied to other pollen families in the future and will thus
\nhelp to make allergenic pollen monitoring more specific
Keywords:
Asthma
;
Atmospheric science
;
Computer science
;
Environmental sciences
;
Plant sciences
;
Transmission light microscopy
Repository Name:
National Museum of Natural History, Netherlands
Type:
info:eu-repo/semantics/article
Format:
application/pdf
Permalink