Automatic Segmentation of Heart Slices to Rapidly Analyze Treatments for Ischemia-Reperfusion Injury

Abstract

Record ID: 187

Award(s): Excellence in Research Mentoring

Program Affiliation: SURF Program (Summer Undergraduate Research Fellowship)

Student Major: Medical Sciences

Project Advisor: Kevin Haworth

Abstract: Ischemia-reperfusion injury is a common outcome associated with cardiovascular disease, the leading cause of death in the world, and many potential therapies are being developed. The Langendorff ex vivo heart preparation allows for testing cardioprotective therapies in controlled environments. After conducting experiments, the hearts are sliced, scanned, and analyzed to determine the effectiveness of the treatment by differentiating viable and infarct tissue. The first step of the process is manual segmentation to identify which pixels of the scanned images correspond to heart tissue versus background. The process is time-consuming and prone to human inconsistency. Therefore, we are exploring an automatic segmentation technique that can improve the speed and consistency of the analysis. We are investigating whether a particular color channel (red, green, or blue) can be used to identify the pixels in the scanned images as part of heart tissue or background in an image. A receiver operator characteristic (ROC) analysis is used to determine how well each color channel differentiates background versus heart tissue which is done for both black and white backgrounds. The gold standard was the manual segmentation method. We have collected data on the effectiveness of these methods by comparing 96 images (6 slices for 16 different hearts). The outcome of this study can improve the speed of the analysis process, while providing an understanding of the accuracy relative to a gold standard. Automating the process of determining viable tissue from infarct tissue will reduce the potential for inconsistency between different individuals manually performing segmentation.