Automatic malaria diagnosis by the use of multispectral contrast imaging
O.K. Bagui1, J.T. Zoueu1*, C. Wählby2
1Laboratoire d’Instrumentation Image et Spectroscopie, INP-HB, DFR-GEE,
B.P 1093 Yamoussoukro, Côte d’Ivoire
2Centre for Image Analysis, Uppsala University and Swedish University of Agricultural Sciences
* Corresponding author. E-mail: email@example.com
Received: 27 May 2014; revised version accepted: 25 November 2014
We report automatic parasitemia measurement strategies for staining-free malaria-infected thin blood smears, by the use of a multimodal and multispectral light emitting diode microscope. Based on the unique optical spectral fingerprint of the individual blood cells in transmission, reflection and scattering modes we count and differentiate healthy and parasitized red blood cells. Our algorithm uses the third principal component of the optical spectrum of each erythrocyte to classify the cells and estimate the percentage of parasitized blood cells. We also discuss the scattering angular dependence of the contrast function as well as the method’s robustness to blood smear age. We show that the method is stable on blood smears 10 hours after sample collection; the spectral shape is not altered considerably. We have also found an optimal scattering angle that maximizes contrast, improving the method performance on spectral images. We were able to detect up to 96.57% of parasitized red blood cells as compared to manual counting, with a sensitivity of up to 93.25%.
Keywords: Automatic parasitemia; Multimodal and multispectral microscope; Optical spectral fingerprint; Principal component; Scattering angular; Contrast function.