Medica l Image Ana lysis & Ar t i f icia l Intel l igence Symposium 2022 Nr. 17: Dr. Amirreza Mahbod, PhD: A Two-Stage U-Net Algorithm for Segmentation of Nuclei in H&E-Stained Tissues Curriculum Vitae: Dr. Amirreza Mahbod obtained his BSc and first MSc degrees in Electrical Engineering from the University of Science and Technology in Iran. He also received his second MSc in Biomedical Engineering from the KTH Royal Institute of Technology, Sweden. He completed his PhD in 2020, where he served as an industrial PhD fellow, working jointly at the Medical University of Vienna and TissueGnostics GmbH. He mainly worked on analysing various tissues in microscopic images for his PhD thesis. From January 2020 to July 2022, he worked as a postdoctoral fellow at the Medical University of Vienna. He joined the Research Center for Medical Image Analysis and Artificial Intelligence at Danube Private University in August 2022, where he continues his research on novel deep learning-based algorithms for medical image analysis. Abstract: A Two-Stage U-Net Algorithm for Segmentation of Nuclei in H&E-Stained Tissues Nuclei segmentation is an important but challenging task in the analysis of hematoxylin and eosin (H&E)-stained tissue sections. While various segmentation methods have been proposed, machine learning-based algorithms and in particular deep learning-based models have been shown to deliver better segmentation performance. In this work, we propose a novel approach to segment touching nuclei in H&E-stained microscopic images using U-Net-based models in two sequential stages. In the first stage, we perform semantic segmentation using a classification U-Net that separates nuclei from the background. In the second stage, the distance map of each nucleus is created using a regression U-Net. The final instance segmentation masks are then created using a watershed algorithm based on the distance maps. Evaluated on a publicly available dataset containing images from various human organs, the proposed algorithm achieves an average aggregate Jaccard index of 56.87%, outperforming several state-of-the-art algorithms applied on the same dataset.
RkJQdWJsaXNoZXIy NTA3MTM=