Blaz Cugmas, Eva Štruc, Inese Bērziņa, Mindaugas Tamošiūnas, Laura Goldberga, Thierry Olivry, Kārlis Zviedris, Roberts Kadiķis, Maksims Ivanovs, Miran Bürmen, Peter Naglič. Automated classification of pollens relevant to veterinary medicine. 2024 IEEE 14th International Conference Nanomaterials: Applications & Properties (NAP), IEEE, 2024.
Bibtex citation:
Bibtex citation:
@inproceedings{17242_2024,
author = {Blaz Cugmas and Eva Štruc and Inese Bērziņa and Mindaugas Tamošiūnas and Laura Goldberga and Thierry Olivry and Kārlis Zviedris and Roberts Kadiķis and Maksims Ivanovs and Miran Bürmen and Peter Naglič},
title = {Automated classification of pollens relevant to veterinary medicine},
journal = {2024 IEEE 14th International Conference Nanomaterials: Applications & Properties (NAP)},
publisher = {IEEE},
year = {2024}
}
author = {Blaz Cugmas and Eva Štruc and Inese Bērziņa and Mindaugas Tamošiūnas and Laura Goldberga and Thierry Olivry and Kārlis Zviedris and Roberts Kadiķis and Maksims Ivanovs and Miran Bürmen and Peter Naglič},
title = {Automated classification of pollens relevant to veterinary medicine},
journal = {2024 IEEE 14th International Conference Nanomaterials: Applications & Properties (NAP)},
publisher = {IEEE},
year = {2024}
}
Abstract: Pollen monitoring is helpful for medical purposes, weather or crop forecasting, and climate change analysis. Several automatic pollen sampling and identification systems based on digital microscopy, elastic light scattering, fluorescence, and digital holography have been proposed as an aid for real-time monitoring. Pollens are important aeroallergens also in veterinary medicine, leading to flares of atopic dermatitis in dogs, cutaneous or respiratory allergy in horses, and feline atopic syndrome. In this study, we tested a machine learning model, MobileNet V3 Large, for pollen classification in whole slide images. We included four pollen types relevant to veterinary medicine: Bermuda (Cynodon dactylon) and Timothy grass (Phleum pratense), silver birch (Betula pendula), and olive tree (Olea europaea). The average classification accuracy was 88 %. Most misidentifications occurred between the two grass pollens. After grouping Bermuda and Timothy grass together, the average classification accuracy increased to 98%. We showed that the machine learning model can be beneficial for the automated identification of pollens relevant to veterinary medicine.