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

Diatoms represent one of the morphologically and taxonomically most diverse groups of microscopic eukaryotes. Light microscopy-based taxonomic identification and enumeration of frustules, the silica shells of these microalgae, is broadly used in aquatic ecology and biomonitoring. One key step in emerging digital variants of such investigations is segmentation, a task that has been addressed before, but usually in manually captured megapixel-sized images of individual diatom cells with a mostly clean background. In this paper, we applied deep learning-based segmentation methods to gigapixel-sized, high-resolution scans of diatom slides with a realistically cluttered background. This setup requires large slide scans to be subdivided into small images (tiles) to apply a segmentation model to them. This subdivision (tiling), when done using a sliding window approach, often leads to cropping relevant objects at the boundaries of individual tiles. We hypothesized that in the case of diatom analysis, reducing the amount of such cropped objects in the training data can improve segmentation performance by allowing for a better discrimination of relevant, intact frustules or valves from small diatom fragments, which are considered irrelevant when counting diatoms. We tested this hypothesis by comparing a standard sliding window / fixed-stride tiling approach with two new approaches we term object-based tile positioning with and without object integrity constraint. With all three tiling approaches, we trained Mask-R-CNN and U-Net models with different amounts of training data and compared their performance. Object-based tiling with object integrity constraint led to an improvement in pixel-based precision by 12-17 percentage points without substantially impairing recall when compared with standard sliding window tiling. We thus propose that training segmentation models with object-based tiling schemes can improve diatom segmentation from large gigapixel-sized images but could potentially also be relevant for other image domains.

Label-Free Raman Microspectroscopy for Identifying Prokaryotic Virocells.

Raman microspectroscopy has been used to thoroughly assess growth dynamics and heterogeneity of prokaryotic cells, yet little is known about how the chemistry of individual cells changes during infection with virulent viruses, resulting in so-called virocells. Here, we investigate biochemical changes of bacterial and archaeal cells of three different species in laboratory cultures before and after addition of their respective viruses using single-cell Raman microspectroscopy. By applying multivariate statistics, we identified significant differences in the spectra of single cells with/without addition of virulent dsRNA phage (phi6) for Pseudomonas syringae. A general ratio of wavenumbers that contributed the greatest differences in the recorded spectra was defined as an indicator for virocells. Based on reference spectra, this difference is likely attributable to an increase in nucleic acid versus protein ratio of virocells. This method also proved successful for identification of Bacillus subtilis cells infected with the double-stranded DNA (dsDNA) phage phi29, displaying a decrease in respective ratio, but failed for archaeal virocells (Methanosarcina mazei with the dsDNA methanosarcina spherical virus) due to autofluorescence. Multivariate and univariate analyses suggest that Raman spectral data of infected cells can also be used to explore the complex biology behind viral infections of bacteria. Using this method, we confirmed the previously described two-stage infection of P. syringae's phi6 and that infection of B. subtilis with phi29 results in a stress response within single cells. We conclude that Raman microspectroscopy is a promising tool for chemical identification of Gram-positive and Gram-negative virocells undergoing infection with virulent DNA or RNA viruses. IMPORTANCE Viruses are highly diverse biological entities shaping many ecosystems across Earth. However, understanding the infection of individual microbial cells and the related biochemical changes remains limited. Using Raman microspectroscopy in conjunction with univariate and multivariate statistics, we established a marker for identification of infected Gram-positive and Gram-negative bacteria. This nondestructive, label-free analytical method at single-cell resolution paves the way for future studies geared towards analyzing virus-host systems of prokaryotes to further understand the complex chemistry and function of virocells.

Epiphytic diatom community structure and richness is determined by macroalgal host and location in the South Shetland Islands (Antarctica).

The marine waters around the South Shetland Islands are paramount in the primary production of this Antarctic ecosystem. With the increasing effects of climate change and the annual retreat of the ice shelf, the importance of macroalgae and their diatom epiphytes in primary production also increases. The relationships and interactions between these organisms have scarcely been studied in Antarctica, and even less in the volcanic ecosystem of Deception Island, which can be seen as a natural proxy of climate change in Antarctica because of its vulcanism, and the open marine system of Livingston Island. In this study we investigated the composition of the diatom communities in the context of their macroalgal hosts and different environmental factors. We used a non-acidic method for diatom digestion, followed by slidescanning and diatom identification by manual annotation through a web-browser-based image annotation platform. Epiphytic diatom species richness was higher on Deception Island as a whole, whereas individual macroalgal specimens harboured richer diatom assemblages on Livingston Island. We hypothesize this a possible result of a higher diversity of ecological niches in the unique volcanic environment of Deception Island. Overall, our study revealed higher species richness and diversity than previous studies of macroalgae-inhabiting diatoms in Antarctica, which could however be the result of the different preparation methodologies used in the different studies, rather than an indication of a higher species richness on Deception Island and Livingston Island than other Antarctic localities.

Deep learning-based diatom taxonomy on virtual slides.

Deep convolutional neural networks are emerging as the state of the art method for supervised classification of images also in the context of taxonomic identification. Different morphologies and imaging technologies applied across organismal groups lead to highly specific image domains, which need customization of deep learning solutions. Here we provide an example using deep convolutional neural networks (CNNs) for taxonomic identification of the morphologically diverse microalgal group of diatoms. Using a combination of high-resolution slide scanning microscopy, web-based collaborative image annotation and diatom-tailored image analysis, we assembled a diatom image database from two Southern Ocean expeditions. We use these data to investigate the effect of CNN architecture, background masking, data set size and possible concept drift upon image classification performance. Surprisingly, VGG16, a relatively old network architecture, showed the best performance and generalizing ability on our images. Different from a previous study, we found that background masking slightly improved performance. In general, training only a classifier on top of convolutional layers pre-trained on extensive, but not domain-specific image data showed surprisingly high performance (F1 scores around 97%) with already relatively few (100-300) examples per class, indicating that domain adaptation to a novel taxonomic group can be feasible with a limited investment of effort.

Quantitative comparison of taxa and taxon concepts in the diatom genus Fragilariopsis: a case study on using slide scanning, multiexpert image annotation, and image analysis in taxonomy1.

Semiautomated methods for microscopic image acquisition, image analysis, and taxonomic identification have repeatedly received attention in diatom analysis. Less well studied is the question whether and how such methods might prove useful for clarifying the delimitation of species that are difficult to separate for human taxonomists. To try to answer this question, three very similar Fragilariopsis species endemic to the Southern Ocean were targeted in this study: F. obliquecostata, F. ritscheri, and F. sublinearis. A set of 501 extended focus depth specimen images were obtained using a standardized, semiautomated microscopic procedure. Twelve diatomists independently identified these specimen images in order to reconcile taxonomic opinions and agree upon a taxonomic gold standard. Using image analyses, we then extracted morphometric features representing taxonomic characters of the target taxa. The discriminating ability of individual morphometric features was tested visually and statistically, and multivariate classification experiments were performed to test the agreement of the quantitatively defined taxa assignments with expert consensus opinion. Beyond an updated differential diagnosis of the studied taxa, our study also shows that automated imaging and image analysis procedures for diatoms are coming close to reaching a broad applicability for routine use.

SHERPA: an image segmentation and outline feature extraction tool for diatoms and other objects.

BACKGROUND: Light microscopic analysis of diatom frustules is widely used both in basic and applied research, notably taxonomy, morphometrics, water quality monitoring and paleo-environmental studies. In these applications, usually large numbers of frustules need to be identified and/or measured. Although there is a need for automation in these applications, and image processing and analysis methods supporting these tasks have previously been developed, they did not become widespread in diatom analysis. While methodological reports for a wide variety of methods for image segmentation, diatom identification and feature extraction are available, no single implementation combining a subset of these into a readily applicable workflow accessible to diatomists exists. RESULTS: The newly developed tool SHERPA offers a versatile image processing workflow focused on the identification and measurement of object outlines, handling all steps from image segmentation over object identification to feature extraction, and providing interactive functions for reviewing and revising results. Special attention was given to ease of use, applicability to a broad range of data and problems, and supporting high throughput analyses with minimal manual intervention. CONCLUSIONS: Tested with several diatom datasets from different sources and of various compositions, SHERPA proved its ability to successfully analyze large amounts of diatom micrographs depicting a broad range of species. SHERPA is unique in combining the following features: application of multiple segmentation methods and selection of the one giving the best result for each individual object; identification of shapes of interest based on outline matching against a template library; quality scoring and ranking of resulting outlines supporting quick quality checking; extraction of a wide range of outline shape descriptors widely used in diatom studies and elsewhere; minimizing the need for, but enabling manual quality control and corrections. Although primarily developed for analyzing images of diatom valves originating from automated microscopy, SHERPA can also be useful for other object detection, segmentation and outline-based identification problems.