%0 Conference Proceedings
%T Detection of leukocytes in intravital video microscopy based on the analysis of Hessian matrix eigenvalues
%D 2015
%A Gregório da Silva, Bruno César,
%A Carvalho-Tavares, Juliana,
%A Ferrari, Ricardo José,
%@affiliation Federal University of São Carlos
%@affiliation Federal University of Minas Gerais
%@affiliation Federal University of São Carlos
%E Papa, João Paulo,
%E Sander, Pedro Vieira,
%E Marroquim, Ricardo Guerra,
%E Farrell, Ryan,
%B Conference on Graphics, Patterns and Images, 28 (SIBGRAPI)
%C Salvador
%8 Aug. 26-29, 2015
%I IEEE Computer Society
%J Los Alamitos
%S Proceedings
%K detection of leukocytes, intravital microscopy, temporal image registration.
%X Detection of rolling and adhered leukocytes in intravital microscopy image sequences is an important task in studies of leukocyte-endothelial interactions in the microcirculation of living small animals under different inflammatory conditions. This procedure is usually performed by visual assessment of the image sequences. However, despite being tedious and time consuming, this procedure is prone to the inter- and intra-observer variability. In this work, we developed an automated computer system for the detection of leukocytes in intravital video microscopy. First, the video frames were processed by the bilateral filter to reduce noise while preserving sharp edges. Then, a demons-based image registration technique was applied to minimize animal motion. Finally, the detection of leukocytes was performed by local analysis of Hessian matrix eigenvalues. Quantitative and qualitative evaluation of the proposed method were conducted by using 220 video frames obtained from an experimental study performed on the brain microvasculature of mice. The proposed method was compared with the template matching technique using precision, recall and F1-Score measures. For the Hessian-based method, the results of precision, recall and F1-Score were, respectively, equal to 0.81, 0.86, and 0.83. For direct comparison, the results obtained for the template matching technique were 0.86, 0.73 and 0.79.
%@language en
%3 manuscript_sibigrapi_2015.pdf