Abstract
Quantitative information is essential to the empirical analysis of biological systems. In many such systems, spatial relations between anatomical structures is of interest, making imaging a valuable data acquisition tool. However, image data can be difficult to analyse quantitatively. Many image processing algorithms are highly sensitive to variations in the image, limiting their current application to fields where sample and image quality may be very high. Here, we develop robust image processing algorithms for extracting structural information from a dataset of high-variance histological images of inflamed liver tissue obtained during necropsies of wild Soay sheep. We demonstrate that features of the data can be measured in a fully automated manner, providing quantitative information which can be readily used in statistical analysis. We show that these methods provide measures that correlate well with a manual, expert operator-led analysis of the same images, that they provide advantages in terms of sampling a wider range of information and that information can be extracted far more quickly than in manual analysis.
Highlights
Imaging is nowadays a valued method for data acquisition in biology [1]
Of the individual inflammatory zones found by our algorithm in widefield images, 98% (724/739) have at least 80% of their surface area in common with operator-defined inflammatory zones; this indicates that most inflammatory zones found by our algorithm are, inflammatory regions, as defined by the operator
Gross pathology of the liver provides essential information to determine the aetiology of pathologic processes contributing to mortality in wild animals experiencing infectious and nutritional challenges [38], including the Soay sheep population studied here [22]
Summary
Imaging is nowadays a valued method for data acquisition in biology [1]. Able to capture form, structure, pattern and texture, imaging data present a potentially rich repertoire of information not always available to other sampling modalities. Prior to the considerable advances in computer vision over the last decade, much analysis of image data required operator input, with most measurements typically taking the form of qualitative descriptions or semi-quantitative scores. Such scoring systems are widely used to extract diagnostic and prognostic information in histopathology, where this type of information has a clear clinical interpretation. These scoring methods are known to contain inconsistencies owing partly to operator interpretation bias [2,3]. Similar difficulties in the extraction of quantitative data from images are prevalent across a number of fields of biology, including ecology [4,5,6,7]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
