In an article published in Scientific Data, researchers presented a reference bioimaging dataset consisting of macroscopic and bright-field microscopic images of phenotypic characters of the European species of the liverwort family of Scapaniaceae.
The authors annotated the reference bioimaging data with machine-actionable metadata using semantics widely recognized in the community to promote data reuse in biodiversity and related study areas. Raw macroscopic and bright-field microscopic imaging data encouraged sound scientific standards through source tracking and provenance. Any contextual image analysis, such as stitching and multi-focus image fusion, was also documented.
For image segmentation and machine learning applied in computational ecology and bioinformatics, the data in the raw macroscopic and bright-field microscopic images were of considerable interest. The findings anticipated that this extensively annotated reference bioimaging dataset would inspire additional research to adhere to the guidelines.
Challenges in Assessing the Phenotypic Characters of Bryophytes
The researchers documented organisms’ diversity at the genetic, physiological, metabolic, morphological, and ecosystem level.
Reference bioimaging data in the field of biodiversity are still highly sparse, despite the widespread use of various bioimaging techniques such as macro and microscopy. This scarcity is particularly prominent for underrepresented and taxonomically challenging groups like bryophytes.
Unlike vascular plants, bryophytes do not have organs that are well-differentiated and shield them from pathogens and environmental exposure. Since bryophytes have evolved unique metabolisms and cell structures, such as oil bodies, their phenotypic characters are frequently cryptic and challenging to distinguish visually.
In Europe, the Scapaniaceae family, which includes 48 different taxa, plays a significant role in environmental adaptations, terpenoid biochemistry, other chemical structures, heavy metal metabolism, and phylogenetics. In general, bryophytes have very few documented traits. Mainly, phenotypic characters to evaluate the variety of functions and forms are understudied in liverworts such as Scapaniaceae.
In biodiversity analysis, reference bioimaging data are critical since they allow the evaluation of images to determine the diversity of phenotypic characters.
Many ecological studies are built on quantifiable phenotypic characters such as biological images. Measurement of molecular structures related to genetics, biotechnology, and molecular pathways can be made qualitatively and quantitatively by phenotyping. Phenotypisation involves recording images of morphological and anatomical characteristics.
In this paper, meta-synthesis techniques that combined various data sources from published case studies offered a great prospect of identifying context dependencies in reference bioimaging research data. Any raw or segmented image in the reference bioimaging dataset was paired with a rich set of expressive and contextual metadata, recording the phenotypic characters to establish provenance, encourage reuse, and ensure reproducibility. The metadata was annotated with semantics widely accepted by the community, enabling machine-actionable data mining.
The authors outlined the methods for creating reference macroscopic and bright-field microscopic images from unprocessed microscopic bioimaging data. They also demonstrated how each image was linked to specific technical and expressive information. Although they linked the macroscopic and bright-field microscopic images to a set of metadata, reference bioimaging lacked practical ontological concepts for connecting specific images to phenotypic characters.
The high-resolution macroscopic and bright-field microscopic images enabled enlarged prints and zooming into macroscopic and bright-field microscopic images to gain essential information. The findings were crucial for species identification and computer-assisted species recognition, computational image analysis, and identification.
Using computational techniques, images with extended field depth were constructed from recordings made at various focal planes. This "focus stacking" method was automated for macroscopy by mounting the camera to the Cognisys StackShot macro rail fixed to a Novoflex macro stand.
Adobe Camera RAW was used to prepare the raw images recorded in CR3-format. The software was also used to perform non-destructive image processing, including field curvature adjustments, chromatic aberration removal, and contrast and brightness enhancements.
Helicon Focus 7.7.5 helped execute multi-focus image fusion on the separate images in the z-stacks.
The algorithms depth map and pyramid were selected, each with a different radius and smoothing setting. The most effective composite images were manually selected and maintained. Affinity Photo 1.10.1's panorama stitching feature was used to piece together multiple images when composite images included specimens larger than the frame.
Any raw image was assigned metadata based on specific file names, such as the (i) species name, (ii) taxonomic rank information, (iii) voucher specimen id, (iv) image collection date, (v) object description comprising the name of the collected phenotypic characters, (vi) the employed objective, (vii) microscope, and (viii) magnification.
Multi-focused image fusion algorithms were performed with various settings on the individual images in stacks to test the technical quality of merged composite images. The best image out of the composite images was selected after manual inspection.
Reference Bioimaging and the Future of Biodiversity and Computational Ecology
The authors presented the fundamentals of producing reference images from unprocessed microscopic bioimaging data. The relationship between individual images and expressive and technical information was also demonstrated.
The study's findings proposed a high-quality reference bioimaging dataset consisting of macroscopic and bright-field microscopic images that illustrated numerous phenotypic characters of European members of the liverwort family, Scapaniaceae.
From a technological standpoint, the reference bioimaging data utilized two main techniques, multi-focus image fusion and image stitching, to dramatically extend the depth of field and boost the resolution of the composite macroscopic and bright-field microscopic images. In this context, the raw data also enabled reuse for fusing images with computational super-resolution.
The reference bioimaging data framework facilitated the incorporation of bioimaging data into other research areas. Through source tracing and provenance, the reuse of data and the extensive metadata description would promote sound scientific practices. The authors believe the proposed framework will inspire further data reuse and meta-synthesis in biodiversity and computational ecology.
Peters, K., König-Ries, B. (2022) Reference bioimaging to assess the phenotypic trait diversity of bryophytes within the family Scapaniaceae. Scientific Data, 9(1). https://www.nature.com/articles/s41597-022-01691-x