The scattering characteristics of the malaria byproduct hemozoin, including its scattering distribution and depolarization, are modeled using Discrete Dipole Approximation (DDA) and compared to those of healthy red blood cells. Scattering (or dark-field) spectroscopy and imaging are used to identify hemozoin in fresh rodent blood samples. A new detection method is proposed and demonstrated using dark-field in conjunction with cross-polarization imaging and spectroscopy. SNRs greater than 50:1 are achieved for hemozoin in fresh blood without the addition of stains or reagents. The potential of such a detection system is discussed.
© 2011 OSA
Malaria remains a devastating disease in many parts of the world. Accurate diagnosis of malaria has become increasingly important as drug resistance spreads and medications become increasingly costly . For more than a century the gold standard for malaria diagnosis has been Giemsa stain microscopy. Microscopy using thick smears can consistently provide accurate counts down to 5-20 infected red blood cells (RBCs) per microliter of blood in a laboratory setting . In addition, thin smear tests can provide effective speciation. However, due to the need for a microscope and a trained biologist to operate it, Giemsa stain microscopy is seldom performed outside of a laboratory, limiting its usefulness for both case management and screening in the developing world.
This lack of point-of-care diagnosis has led to the development of antigen-based rapid diagnostic tests (RDTs). While RDTs can be self-administered, they have a detection limit significantly higher than that of microscopy, suffer from inconsistent performance, and may have problems with shelf stability . In addition, at $0.50-$1.50 per test the cost of RDTs is not significantly cheaper than subsidized antimalarial treatments, providing little monetary advantage to healthcare providers to use the test . The issues of diagnostic cost and performance will become increasingly important in any future eradication campaign, where parasite loads will be increasingly low and test volumes high. Thus there has been a call for new malaria diagnostics combining the convenience of RDTs with the performance of microscopy, while reducing cost per test.
As an alternative to Giemsa stain microscopy, dark-field (DF) microscopy was first used to observe the morphological features of malarial infected RBCs in blood in the 1930s [4–6]. This method images the malarial byproduct hemozoin, which forms within the food vacuole of malaria parasites as they digest the hemoglobin of their host RBC. Hemozoin is crystallized by the parasite to sequester free heme from hemoglobin, which is toxic to the parasite. By imaging hemozoin directly, DF microscopy does not require exogenous contrast agents to diagnose malaria. Despite this, DF microscopy has been used mostly with blood smears, which must be submerged in glycerin to achieve good contrast, providing little advantage in simplicity to Giemsa staining. Presumably because of this and the need for a more complex microscope system, in the past nearly eighty years DF microscopy diagnosis has received little attention, especially related to field diagnosis. Another hemozoin-based microscopy method is cross or orthogonal polarization (xP) imaging. This method has been used since the 1980’s for various applications in malaria research [7,8]. However, this method has also not seen wide use. Both DF and xP microscopy have been effectively used in cell-counting flow cytometry systems for laboratory malaria research. A method known as depolarized side-scatter or depolarized laser light has been used with flow cytometry for malaria diagnosis with varied success [9,10]. The complexity of cytometric systems, however, precludes them from use in wide-spread diagnosis.
In this paper we adapt and integrate reflection-mode DF and xP microscopy into one system (hereafter referred to as DFxP) that can detect hemozoin in fresh blood. Using the system illustrated in Fig. 1(a) , hemozoin was found to exhibit highly efficient scattering versus healthy RBCs at certain angles. By optimizing our system to selectively collect only the light scattered from hemozoin, high SNR was achieved.
This novel DFxP combination has multiple advantages that may permit the development of a simple, effective and field-deployable optical diagnostic system. DFxP, with its high SNR and little dependence on morphology, lends itself to automated detection, and recent advances in high-brightness LEDs support simple, robust and controllable illumination on a portable platform. A further advantage of a detection system requiring only fresh blood is that it eliminates the need for smearing, staining, or diluting, thereby facilitating point-of-care diagnosis. These advantages contrast sharply with Giemsa stain and previously attempted DF and xP microscopy techniques, which require careful sample preparation, more complicated systems, and need a skilled microscopist to accurately discern morphological features to compensate for poor discriminatory capability. The advantages of DFxP are addressed in more detail in section 4.
2.1 Malarial blood
Experiments were performed with fresh rodent blood infected with Plasmodium yoelii, a species of rodent malaria, obtained from Seattle Biomedical Research Institute in Seattle, WA. P. yoelii has been found to have hemozoin of nearly identical morphology to that found in human Plasmodium species . Samples were taken from rodents, treated with heparin to prevent coagulation, and then refrigerated until use within two days, during which time no visible degradation was observed. Where appropriate, results were confirmed with Giemsa-stain microscopy. For spectroscopy, hemozoin was extracted from high-parasitemia blood samples and suspended in water using methods described elsewhere .
2.2 Microscopy and spectroscopy
Wet blood samples were prepared by placing a droplet of fresh infected or control blood on a glass slide with a glass coverslip, resulting in a single layer of erythrocytes. A wide variety of imaging modes were tested to determine their ability to detect hemozoin including DF, xP, DFxP, and bright-field. Both reflected and transmitted light were used for each imaging mode. Reflection mode imaging was performed with a Malvern Morphologi microscope while transmission mode imaging was performed with a Nikon Microphot-FXA microscope. Dry objectives ranging from 4x to 100x (numerical aperture (N.A.) = 0.13-0.9) were used to detect hemozoin. Images were captured using a CCD camera.
Spectra were obtained using the Malvern Morphologi in reflection DF mode with or without crossed polarizers. A Nikon LU Plan 20x objective (N.A. = 0.4) was used. The CCD camera was replaced with a fiber-coupled Ocean Optics S2000 spectrometer and a xenon lamp was used for illumination. A fiber collimator was used to increase collection efficiency of the fiber.
2.3 Numerical aperture experiments
The effects of numerical aperture on hemozoin contrast were isolated using a custom bench-top DFxP microscope (see Fig. 1(b)). This system projected the desired illumination ring from a high-power blue LED (peak wavelength 470 nm) onto the back focal plane of a N.A. = 0.75 20x dry objective, thus utilizing the outer ring of the objective’s top lens as a virtual DF condenser. The DF imaging configuration was completed by adding an iris at an aperture conjugate plane in the imaging path of the microscope to reject the illumination ring. A polarizing beam splitter was used to achieve DFxP imaging. The collection, illumination, and gap apertures could be independently varied (see Fig. 3 ). Samples identical to those used for microscopy were used. Images were analyzed for infected/uninfected RBC contrast by comparing an average of the peak pixel values from several infected RBCs in the image to an average of the peak pixel values of several healthy erythrocytes.
Scattering from hemozoin was modeled using Discrete Dipole Approximation (DDA)  and published dielectric properties of hemozoin . Hemozoin crystals were approximated as 200x200x800 nm3 rectangular prisms suspended in water. Due to the convenient size and shape of the hemozoin crystal, convergence in DDA was observed with as little as 500 dipoles. Because scattering was observed to be highly dependent on hemozoin orientation relative to the k and E vectors of the incident light, the scattering distributions over all possible rotational orientations of the particle were averaged to model the scattering properties of a group of dispersed, randomly-oriented particles. For the scattering properties of RBCs, a CST Microwave Studio finite-element scattering model was used due to their large size relative to the wavelength. RBCs were modeled only aligned parallel to the plane normal to the k-vector.
Figure 2 shows some of the unique scattering properties of hemozoin versus those of healthy RBCs. Figures 2(a) and 2(c) show reflected DF images of both control and infected fresh blood samples using a Nikon LU Plan 50x objective. By comparison with previous literature and Giemsa staining it has been confirmed that the bright white or cyan features in certain cells in the infected sample are hemozoin. In addition to the presence of hemozoin, infected cells are also distinguished by being otherwise dimmer (i.e. less scattering), especially those in the later trophozoite and schizont stages. This is to be expected considering that as the Plasmodium parasite consumes hemoglobin, the index of refraction of the cell drops until it is similar to that of the surrounding plasma .
By experimenting with various DF modalities, it was determined that two aspects of the illumination are critical for obtaining high infected RBC to healthy RBC contrast (hereafter referred to as contrast) in fresh blood. First, reflected mode DF is superior to transmitted mode DF in contrast. Modeling confirmed that this is due to the relatively weak backscattering of red blood cells (Fig. 2(f)) compared to hemozoin (Fig. 2(g)). In addition, reflection mode DF imaging produces fewer false positives than transmission mode DF. In particular, densely crowded healthy RBCs can produce bright spots in transmission DF and xP modes that are nearly indistinguishable from hemozoin in shape, brightness, and color. These were referred to as “crowding errors”. Second, the contrast varies greatly with the illumination and collection apertures of the DF system. This is discussed in more detail later for the special case of DFxP (see Fig. 3).
Adding crossed polarizers to reflected DF mode allows DFxP imaging, which is shown in Figs. 2(b) and 2(d). The scattering distributions for RBCs and hemozoin in this configuration are shown in Figs. 2(h) and 2(i), respectively. In reflection-mode DFxP, only depolarized backscatter can be observed, rendering nearly all features in fresh healthy blood invisible. DFxP has more than double the contrast of DF or xP alone and can achieve SNRs of greater than 50. These observations indicate that light scattered at large angles by hemozoin is more depolarize than light scattered by RBCs at the same angle. In addition, DFxP also has lower occurrence of false positives than any other imaging mode included in this trial and does not produce crowding errors, even from multiple densely packed layers of RBCs.
The DF or scattering spectrum of extracted hemozoin in water was measured and is shown in Fig. 2(e). The calculated curve is of the scattering cross-section determined by DDA, normalized to fit the unitless spectral data. The DF spectrum of hemozoin is characterized by a peak at 670nm, a general increase in scattering at shorter wavelengths, and a second peak between 425 and 450 nm. There is a small difference between the calculated and measured spectra, possibly due to traces of highly absorbing hemoglobin left in the hemozoin extraction. However, over most of the visible range, it appears that the DF spectrum closely matches the calculated values.
The DFxP spectrum was also measured for the extracted hemozoin. The DFxP spectrum retained the peaks at 670 and between 425 and 450nm, however, the increase in scattering at shorter wavelengths is less pronounced. The calculated DFxP curve, representing the depolarized scattering cross-section, reflects these features but shows an overall poorer fit than the DF spectrum. The measured DFxP scattering was approximately 20 times less intense than the DF scattering at 550 nm. Note that the depolarized signal from hemozoin was modeled with an isotropic complex index of refraction, implying that material birefringence is not necessary to describe the depolarizing properties of hemozoin. Indeed, experiments in transmission xP mode show that hemozoin contrast extinguished as the condenser aperture was closed, implying that only scattered light is depolarized and not light directly transmitted through the hemozoin - an observation consistent with our non-birefringent model.
For comparison, the DF and DFxP spectra of healthy blood are included in Fig. 2(e). The DF and DFxP spectra of healthy blood are dominated by the absorption bands of oxyhemoglobin. Due to the low density of infected RBCs in infected blood, it proved difficult to detect hemozoin using spectral measurements alone, as the spectrum is dominated by the oxyhemoglobin features. To observe the hemozoin spectral signature in an infected blood sample, the hemoglobin had to first be degraded by thermal or chemical means.
Using the custom DFxP microscope previously described in section 2.3, the effect of numerical aperture on image contrast was carefully studied. Blue LEDs were used due to the stronger scattering of hemozoin at shorter wavelengths. By selectively varying the three apertures characteristic of dark-field imaging – collection (αC), gap (αG), and illumination (αI) – the optimal alignment was determined. Results are shown in Fig. 3.
In general, contrast increases with increasing αI and αC. The higher isotropy of hemozoin scattering versus healthy RBCs shown in Figs. 2(f)–2(i) explains this result as high N.A. DF alignment preferentially collects large angle, or more isotropic scattering. In addition to αC and αI, αG was also found to be significant as shown in the inset of Fig. 3, with increased αG resulting in greater contrast. Increasing αG increases the rejection of low-angle backscatter from the sample (direct backscatter is always rejected by DF mode). It appears this rejection of low-angle backscatter is responsible for the increase in contrast. Optimal contrast is therefore a compromise between a large αG and a large αC, as increasing one generally decreases the other.
The maximum contrast in Fig. 3 was approximately 10. This experiment was designed to isolate the scattering of infected and healthy RBCs for accurate comparison. By instead optimizing sample preparation for maximum SNR using a thicker sample chamber with multiple stacked layers of RBCs, SNRs of greater than 50 were achieved.
4. Discussion and conclusion
We have shown that imaging fresh blood samples under DF illumination is superior in reflection mode as opposed to transmission mode in both contrast and false positive rate. Experimental results confirm our DDA analysis that contrast between infected and uninfected RBCs is enhanced due to the relatively weak backscatter (as opposed to forward scatter) from the cells compared to that from hemozoin. Furthermore, the DDA model predicts that large-angle scattering events from hemozoin will also depolarize the incident light. Indeed, we observe that DFxP doubles the contrast of DF or xP alone, with SNRs greater than 50. DFxP also maintains contrast with thick samples while other tested methods lose contrast and exhibit more false positives due to multiple scattering events over longer path lengths. We find that spectral measurements are less useful for identifying infected RBCs because absorption of oxyhemoglobin dominates hemozoin scattering. Finally, in order to optimize contrast, we varied the collection, illumination, and gap apertures to reject low-angle backscatter while collecting more large-angle scattering. High contrast is observed even when using a modest collection aperture of 15.1° (N.A. = 0.26).
While reflected mode DF typically utilizes expensive microscopes and specialized objectives for illumination, advances in LED technology now allow for cheap and compact ways to produce DF illumination. Once a suitable DF illumination cone is produced, any low-N.A. objective can be used for imaging. Expanding such a system to perform DFxP imaging requires only the addition of polarizing film and thus enables the identification of hemozoin in thick samples of fresh blood. Sample preparation for DFxP diagnosis could therefore be as simple as a self-administered glucose test, where blood from a finger stick is pulled into a capillary chamber. Furthermore, the high contrast of DFxP potentially enables automated detection. Our results suggest that, with further optimization, the detection method described here offers the potential for high performance, point-of-care malaria diagnosis in the developing world.
The authors would like to thank Seattle Biomedical Research Institute for supplying rodent blood samples. Dave Nash, Barcin Acar, and Emma Mullen supplied valuable support to this project at Intellectual Ventures Laboratories. The authors thank Bill and Melinda Gates for their active support of this work and their sponsorship through the Global Good Fund.
References and links
1. C. Wongsrichanalai, M. J. Barcus, S. Muth, A. Sutamihardja, and W. H. Wernsdorfer, “A review of malaria diagnostic tools: microscopy and rapid diagnostic test (RDT),” Am. J. Trop. Med. Hyg. 77(6Suppl), 119–127 (2007).
2. D. Payne, “Use and limitations of light microscopy for diagnosing malaria at the primary health care level,” Bull. World Health Organ. 66(5), 621–626 (1988). [PubMed]
3. World Health Organization, “Malaria Rapid Diagnostic Test Performance,” (Geneva, 2009).
4. A. Wolter, “Uber die Schnelldiagnose der Malaria mit Hilfe des Dunkelfeldes,” Dermat. Zeitschrift 63(1-2), 69–72 (1932). [CrossRef]
5. G. A. Jamjoom, “Dark-field microscopy for detection of malaria in unstained blood films,” J. Clin. Microbiol. 17(5), 717–721 (1983). [PubMed]
6. B. R. Wood, A. Hermelink, P. Lasch, K. R. Bambery, G. T. Webster, M. A. Khiavi, B. M. Cooke, S. Deed, D. Naumann, and D. McNaughton, “Resonance Raman microscopy in combination with partial dark-field microscopy lights up a new path in malaria diagnostics,” Analyst (Lond.) 134(6), 1119–1125 (2009). [CrossRef]
7. C. Lawrence and J. A. Olson, “Birefringent hemozoin identifies malaria,” Am. J. Clin. Pathol. 86(3), 360–363 (1986). [PubMed]
8. C. Romagosa, C. Menendez, M. R. Ismail, L. Quintó, B. Ferrer, P. L. Alonso, and J. Ordi, “Polarisation microscopy increases the sensitivity of hemozoin and Plasmodium detection in the histological assessment of placental malaria,” Acta Trop. 90(3), 277–284 (2004). [CrossRef]
9. F. P. Josephine and V. Nissapatorn, “Malaria: the value of the automated depolarization analysis,” Southeast Asian J. Trop. Med. Public Health 36(Suppl 4), 68–72 (2005).
10. B. V. Mendelow, C. Lyons, P. Nhlangothi, M. Tana, M. Munster, E. Wypkema, L. Liebowitz, L. Marshall, S. Scott, and T. L. Coetzer, “Automated malaria detection by depolarization of laser light,” Br. J. Haematol. 104(3), 499–503 (1999). [CrossRef]
11. G. S. Noland, N. Briones, and D. J. Sullivan Jr., “The shape and size of hemozoin crystals distinguishes diverse Plasmodium species,” Mol. Biochem. Parasitol. 130(2), 91–99 (2003). [CrossRef]
12. M. D. Green, L. Xiao, and A. A. Lal, “Formation of hydroxyeicosatetraenoic acids from hemozoin-catalyzed oxidation of arachidonic acid,” Mol. Biochem. Parasitol. 83(2), 183–188 (1996). [CrossRef]
13. B. T. Draine and P. J. Flatau, “Discrete-dipole approximation for scattering calculations,” J. Opt. Soc. Am. A 11(4), 1491–1499 (1994). [CrossRef]
14. Y. M. Serebrennikova, J. Patel, and L. H. Garcia-Rubio, “Interpretation of the ultraviolet-visible spectra of malaria parasite Plasmodium falciparum,” Appl. Opt. 49(2), 180–188 (2010). [CrossRef]
15. Y. Park, M. Diez-Silva, G. Popescu, G. Lykotrafitis, W. Choi, M. S. Feld, and S. Suresh, “Refractive index maps and membrane dynamics of human red blood cells parasitized by Plasmodium falciparum,” Proc. Natl. Acad. Sci. U.S.A. 105(37), 13730–13735 (2008). [CrossRef]