Detecção de leucócitos em imagens de vídeo de microscopia intravital usando a técnica de congruência de fase

Abstract

Quantification of the number of rolling and adherent leukocytes presentin the microcirculation of small animals is an important task to elucidate inflamma-tion mechanisms and to assess therapeutic effects of new drugs. In general, leukocytecounting is done visually by an observer (laboratory technician or specialist) using anintravital microscopy image sequence. However, such task is time-consuming and sus-ceptible to errors, due to visual fatigue of the observer and to inter and intra-observervariabilities. In this work a new computational technique is proposed for the automaticdetection of leukocytes in intravital video microscopy. The technique uses a blobnessmeasure, calculated from eigenanalysis of local matrices of second-order momentsof the phase congruency measure, to enhance leukocytes in the images. Detectionof leukocytes is achieved by searching for local maxima in the map of blobness me-asures. Using a set of frames with the leukocyte centroids manually annotated, theresults were evaluated using the F1-measure and the area under the precision-recallcurves (AUCPRs). The results for the proposed technique (F1−measure = 0,791,AUCPR = 0,776) were superior in comparison with the template matching techni-que (F1−measure = 0,746,AUCPR = 0,670).