ArtPOP - Automated RecogniTion of Palynomorphs and Organic sedimentary Particles

The traditional workflow in palynology begins with the removal of rock minerals through acid digestion and heavy liquid separation, followed by mounting the organic residue on a glass slide, and analysing it under a transmitted light microscope. Using the microscope, palynologists manually identify and assign the observed particles to predefined categories within a designated counting area on each slide. Counting typically continues until a target number of particles has been reached (often between 200 to 300).

Beyond the commonly analysed palynomorphs such as pollen, spores, and dinoflagellate cysts, palynological slides may also contain a diverse range of acid resistant organic sedimentary particles, including freshwater algae, phytoclasts, amorphous organic matter, and many others. Examining the full spectrum of these particles is known as palynofacies analysis. It is one of the most powerful methods for reconstructing depositional environments in sedimentary rocks, as it relies on the distribution and relative abundances of these particles.

However, traditional counting methods for palynological and palynofacies analysis present several limitations. The counting area is rarely defined with precision, making it difficult to reproduce analyses. As a result, if any annotations need to be corrected, the entire counting workflow must be repeated. A particularly challenging aspect is the objective estimation of particles such as amorphous organic matter or phytoclasts, which are always fragmented and do not exist as discrete entities. Moreover, identification accuracy can vary substantially between analysts depending on experience, introducing challenges for reproducibility, comparability, and integration across datasets.

Digitizing palynological slides offers a promising opportunity to reduce subjectivity and personal bias by enabling particle annotation directly on high resolution digital images. This approach also supports iterative analysis, allowing annotations to be updated or refined without repeating the microscopy workflow. Through the ArtPOP project, we aim to develop objective, widely applicable annotation tool that enhance the robustness of paleoenvironmental reconstructions and facilitate integration across diverse palynological datasets. In this presentation, we provide an overview of challenges and advantages associated with digitizing the palynological workflow. We also present our preliminary results of the AI-augmented annotation of selected sedimentary particles.