Journal of Spectral Imaging,   Volume 9   Article ID a10   (2020)

Peer reviewed Paper

Detection and segmentation of erythrocytes in multispectral label-free blood smear images for automatic cell counting

  • Solange Doumun
  • Sophie Dabo
  • Jérémie Zoueu
Université Lille, CNRS, UMR 8524-Laboratoire Paul Painlevé, INRIA-MODAL, F-59000 Lille, France
[email protected]
 https://orcid.org/0000-0002-4000-6752
 Search for papers by this author
Laboratoire d’Instrumentation, Image et Spectroscopie, Institut National Polytechnique, Félix Houphouët-Boigny, BP 1013, Yamoussoukro, Côte d’Ivoire
[email protected]
 https://orcid.org/0000-0003-4588-1641
 Search for papers by this author
 Corresponding Author
Université Lille, CNRS, UMR 8524-Laboratoire Paul Painlevé, INRIA-MODAL, F-59000 Lille, France and Laboratoire d’Instrumentation, Image et Spectroscopie, Institut National Polytechnique, Félix Houphouët-Boigny, BP 1013, Yamoussoukro, Côte d’Ivoire
[email protected]
 https://orcid.org/0000-0001-7049-1499
 Search for papers by this author

In this work we propose an efficient approach to image segmentation for multispectral images of unstained blood films and automatic counting of erythrocytes. Our method takes advantage of Beer–Lambert’s law by using, first, a statistical standardisation equation applied to transmittance images, followed by the local adaptive threshold to detect the blood cells and hysteresis contour closing to obtain the complete blood cell boundaries, and finally the watershed algorithm is used. With this method, image pre-processing is not required, which leads to time savings. We obtained the following results that show that our technique is effective, efficient and fast: Precision of 98.47 % and Recall of 98.23 %, a degree of precision (F-Measurement) of 98.34 % and an Accuracy of 96.75 %.

Keywords: multispectral imaging, segmentation, malaria, automatic diagnosis, image analysis

Metrics

Downloads:

43

Full-Text HTML Views:

1

Abstract Views:

218