Simulation Palynologists for Pollinosis Prevention: A Progressive Learning of Pollen Localization and Classification for Whole Slide Images

Abstract

Existing API approaches usually independently leverage detection or classification models to distinguish allergic pollens from Whole Slide Images (WSIs). However, palynologists tend to identify pollen grains in a progressive learning manner instead of the above one-stage straightforward way. They generally focus on two pivotal problems during pollen identification. (1)Localization: where are the pollen grains located? (2) Classification: which categories do these pollen grains belong to? To perfectly mimic the manual observation process of the palynologists, we propose a progressive method integrating pollen localization and classification to achieve allergic pollen identification from WSIs. Specifically, data preprocessing is first used to cut WSIs into specific patches and filter out blank background patches. Subsequently, we present the multi-scale detection model to locate coarse-grained pollen regions (targeting at “pollen localization problem”) and the multi-classifiers combination to determine the fine-grained category of allergic pollens (targeting at “pollen classification problem”). Extensive experimental results have demonstrated the feasibility and effectiveness of our proposed method.

A user‐friendly method to get automated pollen analysis from environmental samples

“…Nonetheless, when detection was performed jointly with classification, grains with rarer and irregular morphologies were also assigned a higher number of falsely detected debris, thus partially compensating errors. When classification is conducted separately from detection, debris falsely detected are often processed with dedicated classes or processes (Crouzy et al ., 2022; Zhao et al ., 2022), and pollen grains left undetected are usually not considered in the final assessment, which may generate biases. The proposed method, based on an algorithm jointly conducting detection and classification, thus integrates detection errors through both stages, which effectively mitigates both types of detection errors.…”

Weakly Supervised Collaborative Learning for Airborne Pollen Segmentation and Classification from SEM Images

“…LM images are considered as a more routine method due to their simple deployment and short obtaining process. Inspired by advanced computer vision technology, LM-based approaches have been widely used for automatically recognizing airborne pollens [ 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 ]. For example, Li et al [ 11 ] extracted a variety of feature descriptors, including morphological feature descriptors, Fourier descriptors, and Haralick texture feature descriptors.…”

“…Wang et al [ 14 ] presented an improved detector incorporating a self-attention mechanism to distinguish pollen and the background in LM images. Zhao et al [ 15 ] proposed a novel pollen identification framework in a progressive manner, which perfectly mimics the manual observation process of the palynologist. However, in practice, we found that LM images suffer from the following disadvantages: (1) the detection optics used to collect the light source in LM can only focus on a fixed distance, which results in some pollen features having out-of-focus blurring; (2) all the details of the pollen grains fail to be completely captured (especially texture features) since this is particularly hindered by the low resolution of the LM scanner (usually 40× or 20×).…”