Explainable AI for unveiling deep learning pollen classification model based on fusion of scattered light patterns...

Area of Science:

• Environmental Science
• Data Science
• Biotechnology

Background:

• Pollen monitoring is increasingly data-intensive with real-time detectors for airborne pollen classification.
• Deep learning models are crucial for accurate pollen classification but often function as black boxes.

Purpose of the Study:

• To develop an explainable artificial intelligence (xAI) framework for a deep learning pollen classification model.
• To understand the decision-making process of deep learning models in pollen identification.

Main Methods:

• Utilized data from a single particle detector (Rapid-E) capturing optical fingerprints (scattered light, laser-induced fluorescence).
• Integrated three data modalities: scattering, fluorescence spectrum, and fluorescence lifetime.
• Applied explainable AI (xAI) methodologies to a deep learning model trained on this multi-modal data.

Main Results:

• The study presents the first application of xAI to a pollen classification model.
• Identified key features contributing to the classification of specific pollen types.
• Compared extracted feature knowledge with domain expertise and reference data (pollen size, shape, spectrofluorometer measurements).

Conclusions:

• The developed xAI framework provides insights into the 'black-box' deep learning model for pollen classification.
• This approach enhances the reliability and interpretability of automated pollen monitoring systems.
• The findings support the integration of explainable AI in environmental sensing and biodiversity studies.