P517Automated segmentation of gap junctions using confocal fluorescent microscopy

Abstract

Introduction: Fluorescent microscope techniques are frequently used to measure the intensity and the cellular distribution of connexin 43 (Cx43), which is the main gap junction (GJ) forming protein in the heart. In our previous study, a fully automated algorithm was developed for the segmentation and quantification of GJs at the longitudinally oriented myocardial tissue sections. This segmentation method detects separately the sarcolemma and the GJs through which the 'polar' GJs (intensities at the intercalated discs) and ’lateral‘ GJs (signals at the lateral membrane) can be distinguished. Purpose: We have now developed new algorithms to identify GJs not only in the longitudinally but also in the transversely oriented sections and to segment 'polar' GJs without the need of the sarcolemmal staining. Methods and results: Longitudinal and transversal myocardial sections were obtained and were separately labeled for the sarcolemma and Cx43. Furthermore, longitudinal sections were also labeled for Cx43 and for other proteins of the intercalated disc such as ZO1 which is a known interacting partner of Cx43. The fluorescent signals were acquired simultaneously into separate channels using a confocal laser-scanning microscope. After image enhancement, automatic segmentation was applied on the transverse sections and different masks were generated; the cell membrane mask represents the sarcolemma but the artifacts or damaged cells were eliminated during image processing and only the intercalated discs were identified. Using this mask the Cx43 signals can be identified and the intensity as well as the distribution of GJs can be measured on the en face oriented intercalated discs. In the longitudinal sections the 'polar' intensities were separately measured and the co-localization pattern was identified for Cx43 and ZO1 based on our previously developed and modified line tracking algorithm. When we compared these algorithms with other image analyzer (ImageJ v1.47), the same densities and distribution patterns were obtained for Cx43 and ZO1. Discussion: Detection of GJs on the fluorescent images is time-consuming and usually manually performed, because accurate object identification is often not possible with the implementation of traditional segmentation methods (such as edge detection, region growing or histogram-based methods, watershed transformation, model based segmentation etc.). We conclude that these new algorithms provide an automatic, accurate, fast and objective segmentation method for identification of variously oriented GJs.