Fully Convolutional Neural Network for Detection and Counting of Diatoms on Coatings after Short-Term Field Exposure

Abstract

While the use of deep learning is a valuable technology for automatic detection systems for medical data and images, the biofouling community is still lacking an analytical tool for the detection and counting of diatoms on samples after short-term field exposure. In this work, a fully convolutional neural network was implemented as a fast and simple approach to detect diatoms on two-channel (fluorescence and phase-contrast) microscopy images by predicting bounding boxes. The developed approach performs well with only a small number of trainable parameters and a F1 score of 0.82. Counting diatoms was evaluated on a data set of 600 microscopy images of three different surface chemistries (hydrophilic and hydrophobic) and is very similar to counting by humans while demanding only a fraction of the analysis time.

Versatile Microfluidic Platform for Automated Live-Cell Hyperspectral Imaging Applied to Cold Climate Cyanobacterial Biofilms

“…This is an important distinction over previous nonmicrofluidic machine learning studies, which focused on monospecies, , as the interaction between subpopulations of different organisms is a key environmental factor for study, or mixed species with distinct morphological characteristics . The challenges in applying machine learning to mixed-species samples is demonstrated by other studies, which have produced variable results even when acquiring training data from multimodal microscopy. , Moreover, characterization of biofilms is critical for environmental samples, but these samples preclude the use of high-throughput techniques such as flow cytometry . The goal is to use this platform to identify individual cyanobacterial species and even their phenotypes with high accuracy from complex environmental samples ( e.g.…”

“…47 The challenges in applying machine learning to mixed-species samples is demonstrated by other studies, which have produced variable results even when acquiring training data from multimodal microscopy. 48,49 Moreover, characterization of biofilms is critical for environmental samples, but these samples preclude the use of highthroughput techniques such as flow cytometry. 50 The goal is to use this platform to identify individual cyanobacterial species and even their phenotypes with high accuracy from complex environmental samples (e.g., containing gas bubbles and ECM), unlike a recent machine learning approach that achieves unclassified detection among debris.…”

Improving deep learning-based segmentation of diatoms in gigapixel-sized virtual slides by object-based tile positioning and object integrity constraint

“…This has led to extensive research on digitalization and automation of the light microscopy workflow (4,5). The main methodological steps in this context include large-scale automated light microscopic image acquisition at high optical resolution (6)(7)(8)(9)(10)(11)(12); localization of diatom frustules / valves in images (13)(14)(15)(16)(17)(18)(19); extraction of morphometric descriptors from image patches . CC-BY 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.…”

“…This has led to extensive research on digitalization and automation of the light microscopy workflow (4,5). The main methodological steps in this context include large-scale automated light microscopic image acquisition at high optical resolution (6)(7)(8)(9)(10)(11)(12); localization of diatom frustules / valves in images (13)(14)(15)(16)(17)(18)(19); extraction of morphometric descriptors from image patches depicting diatom frustules / valves (12,13,(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30); and taxonomic identification, i.e., labelling of such image patches with a taxonomic name (26,(31)(32)(33). One of the most promising approaches towards large-scale light microscopic image acquisition from standard diatom microscopy slides is high-resolution slide scanning.…”

“…This has led to extensive research on digitalization and automation of the light microscopy workflow (4, 5). The main methodological steps in this context include large-scale automated light microscopic image acquisition at high optical resolution (6-12); localization of diatom frustules / valves in images (13-19); extraction of morphometric descriptors from image patches depicting diatom frustules / valves (12, 13, 20-30); and taxonomic identification, i.e., labelling of such image patches with a taxonomic name (26, 31-33).…”