Date Published: March 11, 2019
Publisher: Public Library of Science
Author(s): Dean Philip John Kavanagh, Meurig Thomas Gallagher, Neena Kalia, Yuhua Zhang.
The ability to image biological tissues is critical to our understanding of a range of systems and processes. In the case of in situ living tissue, such imaging is hampered by the innate mechanical properties of the tissue. In many cases, this provides challenges in how to process large amounts of image data which may contain aberrations from movement. Generally, current tools require the provision of reference images and are unable to maintain temporal correlations within an image set. Here, we describe a tool–Tify–which can accurately predict a numerical quality score versus human scoring and can analyse image sets in a manner that allows the maintenance of temporal relationships. The tool uses regression-based techniques to link image statistics to image quality based on user provided scores from a sample of images. Scores calculated by the software correlate strongly with the scores provided by human users. We identified that, in most cases, the software requires users to score between 20–30 frames in order to be able to accurately calculate the remaining images. Importantly, our results suggest that the software can use coefficients generated from consolidated image sets to process images without the need for additional manual scoring. Finally, the tool is able to use a frame windowing technique to identify the highest quality frame from a moving window, thus retaining macro-chronological connections between frames. In summary, Tify is able to successfully predict the quality of images in an image set based on a small number of sample scores provided by end-users. This software has the potential to improve the effectiveness of biological imaging techniques where motion artefacts, even in the presence of stabilisation, pose a significant problem.
The microcirculation is the primary site of gas and nutrient exchange and also provides the means for the trafficking of various cellular effectors to tissues upon demand. Upon infection, or other deleterious stimuli, circulating inflammatory cells firstly adhere to microvessels and then migrate through them towards damaged areas. Understanding these adhesive events is critical to managing diseases of the microcirculation. Fluorescent intravital microscopy (IVM) is an effective experimental tool allowing visualisation of various microvascular beds in vivo [1–6]. However, due to their anatomical location or normal physiological function, some tissues are in a state of constant movement which creates a significant barrier to imaging. Indeed, the beating nature, and close anatomical proximity to the respiratory movement of the lungs, has meant in vivo imaging of the heart has remained challenging and elusive. To minimize motion artefacts, several stabilisation techniques have been designed that physically constrain the tissue [1–3]. However, although the heart can be held still by attaching a tissue stabiliser, contraction of the myocardium itself within the window of the stabiliser cannot be fully eliminated. Recent advances in microscope and camera technology have allowed investigators to capture dynamic events at increasingly faster frame rates in such organs. However, the data generated are normally larger, with poor quality and out-of-focus frames interspersed throughout the recorded set. As a result, these data sets consist of both visually usable/stable and unusable/unstable frames. While the capture of stable frames allows for analysis of cell trafficking and recruitment, the inclusion of unstable and out-of-focus frames in the video cause problems in both processing and analysis from these data sets.
Intravital microscopy has become an invaluable tool for the monitoring of events in the microcirculation. While some tissues lend themselves well to intravital imaging[13–15], others have proven more difficult to image. In most cases, this is due to issues concerning access, motion and stabilisation. For instance, the lungs and heart have both proven difficult to image due to their anatomical location and respiratory / contractile movements (this is also a common problem for other non-intravital imaging techniques). While novel stabilisation techniques have alleviated some of the problems caused by tissue motion, in most cases these techniques do not fully abolish motion artefacts. Although it is possible to synchronise imaging with physiological triggers (e.g. ECG or respiratory gating), this technology is not available to all laboratories. In these instances, post-hoc image analysis is preferable. Post-hoc image analysis primarily consists of the rejection of frames which do not contribute to analysis (e.g. blurred, out of focus) and the registration of images which are subject to small motion artefacts. In this manuscript, we describe one such post-hoc processing tool which we have named Tify. This package is able to assign scores to image frames based on their calculated quality, on the basis that the user provides a small amount of input data (manual quality score) which can be associated with image statistics using multiple linear regression. The resultant scores can then be used to exclude or include images based on their calculated quality scores.