Automated Quantification of White Blood Cells in Light Microscopy Muscle Images: Segmentation Augmented by CNN
Document Type
Conference Proceeding
Publication Date
8-27-2018
Publication Title
2nd International Conference on Vision, Image and Signal Processing
Volume
2018
First page number:
1
Last page number:
7
Abstract
White blood cells (WBCs) play an important role in the muscle recovery process. Detection and quantification of WBC expressions in light microscopy images captured at different time points after injury deliver valuable information about underlying processes. In this paper, an optimized CNN architecture is designed for classifying CD68 macrophages in 10x light microscopy images of injured muscle cross-sections. Based on the CNN classification results, hybrid masks are generated to post-process the segmentation results obtained by the LIOtsu thresholding method as a step towards extracting and quantifying CD68-positive macrophages. The segmentation is completed by the earlier designed LIOtsu thresholding method. The experimental results confirm that a high accuracy of classification is achieved by the proposed CNN architecture and high performance of quantification of CD68-positive macrophages is achieved by the LIOtsu thresholding method, augmented by CNN.
Keywords
CD68; Convolutional neural network; Muscle; White blood cell
Disciplines
Electrical and Computer Engineering
Language
English
Repository Citation
Jiao, Y.,
Schneider, B. S.,
Regentova, E.,
Yang, M.
(2018).
Automated Quantification of White Blood Cells in Light Microscopy Muscle Images: Segmentation Augmented by CNN.
2nd International Conference on Vision, Image and Signal Processing, 2018
1-7.
http://dx.doi.org/10.1145/3271553.3271570