We initially tested the effectiveness of the presented cytometry approach by imaging various cells and microparticles on a chip. Figures 1–3 in the Supporting Information
illustrate the holographic imaging results of a heterogeneous solution that contains RBCs, yeast cells (S. pombe
), and 3-μ
m-, and 10-μ
m-sized particles all imaged using the lens-free holographic imaging setup of . For these images, we utilized L
= 10 cm, H
= 0.7 mm, an aperture size of D
= 50 μ
m, and a spatially incoherent source (from a xenon-lamp-based monochromator) at 550 nm with a full width at half maximum (fwhm) spectral bandwidth of ~15 nm (see ). As illustrated in these figures, the raw holographic signatures of cells and particles exhibit unique oscillations that are characteristic of each cell/particle type, based on its size, shape, subcellular elements, and refractive index. As we have already demonstrated in an earlier study,8
using this two-dimensional (2D) amplitude hologram, it is feasible to distinguish different cells from each other using pattern analysis techniques (see, e.g., Figure 3 in the Supporting Information
). More importantly, the amplitudes of these cell holograms also permit digital reconstruction of microscopic images of the cells, as well as their phase signatures, complementing the cytometric information that is embedded in the cell hologram texture. (See Figures 1–3 in the Supporting Information
and ref 9
for further details.)
To demonstrate and quantify the cytometry performance of our lens-free holographic setup (shown in ), we imaged whole blood samples at various dilution levels, spanning <50 000 cells/μ
L up to ~0.4 × 106
L. and , as well as Figure 4 in the Supporting Information
, illustrate the lens-free RBC hologram amplitudes, as well as their reconstructed images for various cell density levels, which clearly indicate the strength of the digital holographic reconstruction to handle highly dense cell solutions. Note that the same holographic reconstruction process also permits digital separation of cell holograms from each other. Considering , where all of the lens-free RBC holograms are completely overlapping (because of a high cell density of ~0.4 × 106
L), using the complementary information of the reconstructed cell images, one can individually separate the 2D holographic signature (both phase and amplitude) of any given cell within the hologram crowd, as also illustrated in Figure 5 in the Supporting Information
. This duality between cell hologram texture and the digitally reconstructed cell images is especially useful for characterization of dense cell solutions such as whole blood samples.
Figure 2 Measured cell hologram amplitudes at an RBC concentration of 102 000 cells/μL, using the setup of ; L = 10 cm, H = 0.7 mm, and D = 50 μm. In this case, cell shadow overlap probability on the chip is quite high for the raw holographic (more ...)
Figure 3 Same as in , except for an RBC concentration of ~394 000 cells/μL. At such high cell densities, the individual holographic signatures of the cells (as you would normally find in, e.g., ) get completely washed out by a noise-like (more ...)
Next, for each dilution level, we measured the density of RBCs based on automated counting of digitally reconstructed cell images (see the Experimental Methods section for details). The counting results () revealed that the reconstructed cell images yielded an absolute error rate of <5%, up to a cell density of ~0.4 × 106 cells/μL, when compared to counting results of the same FOV obtained using a conventional lens-based microscope. To further validate our holographic characterization results, the inset in also illustrates a comparison of RBC volume histogram that is estimated based on the same digital reconstructions against a commercially available Coulter Counter, which showed a good fit to our results. In these cell volume calculations, we assumed that RBCs are phase-only objects with an average refractive index of ~1.4. Under these assumptions, the thickness of each RBC becomes proportional to its recovered and unwrapped optical phase, and the base area of each RBC can be estimated by simple thresholding of the recovered phase images. The cell volume of each RBC was then estimated by the product of its thickness and base area.
Figure 4 Automated cell counting accuracy of the lens-free imaging technique is illustrated at various RBC densities, ranging from <50 000 cells/μL up to ~0.4 × 106 cells/μL. For each dilution level, at least 500 cells were automatically (more ...)
In addition to automated RBC count and cell volume measurements, the hemoglobin concentration (HGB) of blood samples was also measured using our lens-less imaging platform with a simple modification of the experimental setup, as shown in . According to the Beer–Lambert model,14
the amount of the photon transmission through a cuvette containing hemolyzed RBCs decreases exponentially with the hemoglobin concentration and the path length of the incident light. Based on this model, using a cuvette thickness of ~8.9 mm and an illumination wavelength of ~420 nm, we measured the HGB concentration of six different patients’ blood samples, using lens-free transmission images. Our results were also compared against a commercially available hematology analyzer (Model Coulter LH750, Beckman Coulter). As described in , the average light intensity collected from ~18K pixels of each lens-free image showed a good fit to the Beer–Lambert model,14
yielding a maximum error of ~2.9%, compared to the hematology analyzer (refer to the Experimental Methods section for details).
Figure 5 Hemoglobin (HGB) measurement results for six different patients’ whole blood samples, using lens-free on-chip imaging, are compared to the results of a commercially available hematology analyzer (Model Coulter LH750, Beckman Coulter). Average (more ...)
Using the same lensless holographic microscopy platform, we also tested the imaging of WBCs in diluted whole blood samples. For this purpose, we used an aqueous staining protocol to label the WBCs (detailed in the Experimental Methods section). Holographic imaging results of such a whole blood solution are illustrated in . For comparison purposes, in the same sample solution, we also included nonstained WBCs, as indicated in with an arrow. Notice that the raw holographic signatures of WBCs and their reconstructed phase holograms are uniquely different from other blood cells, as illustrated in , respectively. illustrates the reconstructed microscopic image of the same field of view, which matches very well with the 40× objective-lens image of the same region, clearly showing the stained and the nonstained WBCs.
Figure 6 Holographic imaging results of WBCs within a diluted whole blood sample are illustrated. The sample contained stained WBCs (see the Experimental Methods section) and, for comparison purposes, in the same solution, we also included nonstained WBCs, as (more ...)
Next, using our holographic imaging platform, we automatically counted the number of WBCs from a patient’s blood sample that is already characterized by a Coulter Counter. For this experiment, we diluted the whole blood sample by factors of 8, 12, and 16 to create three different batches from the same source. For each one of these batches, we took 24 samples (a total of 72 samples) and imaged them for automated counting of the WBCs using the setup of . The counting results are summarized in , as a function of the whole blood dilution factor. For 8× and 12× dilution levels, we have counted, on average, ~119 and ~78 WBCs per each holographic image, whereas for the 16× dilution level, this number dropped to a mean count of ~52 WBCs per image. The standard deviations of the WBC counts per image for these three dilution factors were 16.55, 17.70, and 8.42, respectively. Based on our counts, for dilution factors of 8, 12, and 16, we estimated the original WBC density of the patient to be 4875, 4805, and 4251 WBCs/μL, respectively. These results agree well with the Coulter Counter result, which showed a density of 4700 WBCs/μL for the same patient. Except for the 16× dilution level, our results indicate a mean error of <3.72% for the WBC count, and for the 16× dilution level, our error increased to <10%.
Figure 7 Automated WBC count results are demonstrated using the lens-free holographic imaging platform shown in , as a function of the whole blood dilution factor (8×, 12×, and 16×). For each dilution level, 24 samples were imaged (more ...)
The above illustrated results (–) and – were all imaged while the cells were floating within a solution. We also tested the performance of the presented on-chip cytometry scheme on blood smear samples that are prepared in accordance with conventional staining protocols (see the Experimental Methods section). , and Figure 6 in the Supporting Information
, illustrate the holographic imaging results of these blood smear samples, focusing specifically on the holographic signatures (both amplitude and phase) and the reconstructed microscopic images of various blood cells such as granulocytes, lymphocytes, monocytes, platelets, and RBCs. These results, and their comparison to 40× objective-lens microscope images (NA = 0.6), indicate that the reconstructed holographic images can be used to discriminate subcellular differences among three major types of white blood cells (i.e., granulocytes, lymphocytes, and monocytes), and have the potential to provide three-part differential WBC analysis within the same lens-free digital platform. Note also that some of the platelets in these images are located at a different depth of field than the WBC, which is why they do not appear in some of the reconstructed images. Figure 7 in the Supporting Information
illustrates this fact by digitally zooming to another depth layer to bring the signatures of the platelets back, while defocusing the granulocyte image. This highlights the well-known digital zooming capability of holography, which is another major advantage, when compared to conventional optical microscopes for imaging and characterization of samples in three dimensions.15,16
Figure 8 Holographic imaging results of a blood smear sample, illustrating the lens-free shadow signatures of three major types of WBCs (i.e., granulocytes, lymphocytes, and monocytes) and platelets. Different than , these stained cells are imaged in a (more ...)
All the cell holograms that are discussed so far illustrate the complementary nature of the hologram texture to the reconstructed microscopic images, both for phase and amplitude domains. The textural signature of each cell hologram, before a holographic reconstruction is performed, can reveal important variations among different cell types such as a granulocyte versus lymphocyte (see, e.g., ). This is also an important source of digital information that may potentially permit diagnosis of an infectious disease (such as malaria), based on, e.g., inspection of the infected RBC hologram textures and detection of textural asymmetries against a library of healthy blood cells. In addition, this pattern analysis approach (further discussed in the Experimental Methods section) may simplify the computational complexity to significantly improve the speed of diagnosis. However, such a pattern-matching approach would also suffer from spatial overlapping among cell holograms and, therefore, could significantly benefit from holographic reconstruction9
(see e.g., –) to better handle highly dense cell solutions such as whole blood samples.
The total field-of-view in these earlier images is ~24 mm2
, which is ~10 fold larger than the imaging field-of-view of a typical 10× objective lens (see Figure 1 in the Supporting Information
). A much larger FOV can also be obtained using other types of sensor arrays. For instance, the results in illustrate an imaging FOV of ~18 cm2
, which offers a >500-fold larger FOV, when compared to an optical microscope (e.g., 10× objective lens). This figure is obtained using the setup of under ~600 nm illumination. also illustrate the recovered phase holograms and the corresponding reconstruction images of RBCs for a zoomed region of interest over this 18 cm2
FOV. Note that, in this case, the success of the phase recovery step is quite evident in the comparison of . The tradeoff to achieve a larger FOV is a reduction in spatial resolution that is primarily affected by the large pixel size at the sensor array,9,11
which, in this case, is 6.8 μ
m for the sensor-array utilized in .
Figure 9 Ultrawide FOV imaging of RBCs (over 18 cm2) using the holographic imaging platform of is illustrated. This constitutes a >500-fold larger FOV, when compared to an optical microscope using a 10× objective lens. Panel (b) illustrates (more ...)