In order to demonstrate the proof of concept of our cell-phone-based optofluidic cytometry platform, white blood cells (WBCs) in human whole blood samples were chosen as our model system. White blood cells density in whole blood is routinely tested for clinical diagnosis of various diseases, including infections, leukemia, HIV, and bone marrow deficiencies.17
Various point-of-care-based hematology analyzers have also been developed for WBC counting.18–29
To test our flow cytometry platform on the cell phone, white blood cells in fresh whole blood samples were labeled with SYTO16 fluorescent dyes and diluted as described in the Experimental Methods section without
lysing red blood cells. These 10× diluted and labeled whole blood samples were then flushed through our microfluidic chamber using a syringe pump at a flow rate of ~1 μ
L/min, and a video of the fluorescent emission arising from the labeled WBCs was recorded continuously as shown in the supplementary movie in the Supporting Information
. As illustrated in this supplementary movie
, our custom-designed video analysis software defines a cascade of five counters with a separation distance of ~270 μ
m from each other; and these digital counters dynamically monitor the number of WBCs that go through the microchannel. In order to improve our cell-counting accuracy, the program counted the number of WBCs that passed through each counter line over a period of 210 frames (i.e., ~30 s) and we averaged the number of the counted WBCs from these five independent counters (see the supplementary movie in the Supporting Information
) to get the number of WBCs within this 30 s time frame. On the basis of the volume flow rate and the average number of the counted WBCs, we can calculate the WBC density in the blood sample within this time frame (i.e., 30 s). The same process was repeated for 5–6 min with continuous blood flow, and the WBCs density for each sample was plotted as a function of time as shown in , parts A and B. The average WBC density for each sample was finally calculated by further averaging each one of these dynamic curves over 5–6 min, and our results were compared to the standard test results obtained from a commercially available hematology analyzer (Sysmex KX-21N). As shown in , parts A and B, our WBC density results matched well with the standard test results with <5% error. To further evaluate our cell-phone-based imaging cytometry platform and its counting accuracy, we imaged 12 different patients’ blood sample. Each sample was imaged for 5–6 min, and average WBC density of the whole blood was estimated by processing its corresponding fluorescent video captured by our cell-phone cytometer. The cell-phone-based imaging cytometry results were compared to the parallel results obtained from the Sysmex KX- 21N hematology analyzer. compares the WBC densities obtained using our cell-phone imaging cytometer to the standard results obtained with the Sysmex KX-21N hematology analyzer, which showed a good correlation to our measurements. For these 12 patients’ blood samples (with WBC densities ranging from ~4000 to 8000 μ
), we obtained a correlation coefficient of ~0.93 between the two methods. In addition to this, we also performed Bland–Altman analysis30
on our results (see ), which shows a bias of −339 cells μ
with 95% limits of agreement of −1026 and 347 cells μ
for a wide range of WBC concentrations. This negative bias indicates that our cell-phone cytometry platform is undercounting the WBCs, which may be caused, e.g., by the partial loss of cells within the microfluidic channel and tubing.31
Figure 2 Automated WBC-counting results obtained using our cell-phone-based imaging flow cytometer are demonstrated for two different patients’ whole blood: (A) for a lower WBC density sample (5000 cells μL−1) and (B) for a higher WBC density (more ...)
Figure 3 (A) Comparison of WBC density measurement results obtained with our cell-phone-based imaging flow cytometer is provided against the results of a commercially available hematology analyzer (Sysmex KX-21N) for 12 different patients. A linear regression (more ...)
Besides counting accuracy, throughput is also an important parameter for an imaging cytometer. The throughput in our flow cytometry system is mainly determined by the cell phone’s camera frame rate. In our current implementation, the camera has a relatively slow frame rate of ~7 fps. To further increase the throughput, we can use a cell-phone camera with a higher frame rate, e.g., LG Dare VX9700, which can achieve a frame rate of ~120 fps. This could potentially further improve our flow rate and thus the counting throughput by, e.g., >15-fold, which would reduce the imaging time for, e.g., a whole blood sample to <20 s per test.
We also tested the spatial resolution of our optofluidic design by imaging static fluorescent objects. By changing the external lens of our attachment to the cell phone to a focal length of ~0.6 mm, a geometrical magnification of ~7.8× was achieved, such that the effective pixel size at the sample plane was <0.3 μm. The system spatial resolution was characterized using green fluorescent beads with various sizes including 4, 2, and 1 μm. , top row, illustrates the imaging performance of our cell-phone-based optofluidic fluorescent microscope for several sets of beads. For comparison purposes, the same beads were also imaged by a conventional benchtop fluorescent microscope using a 40× (NA = 0.65) microscope objective as shown in , bottom row. On the basis of , parts B-1 and C-1, we can easily resolve two fluorescent beads that are separated by 4 μm (center to center). In , part E-1, two beads with a center-to-center distance of 2 μm are also successfully resolved by our cell-phone fluorescent microscope. This spatial resolution level is also validated through cross-sectional profiles of isolated 1 μm fluorescent particles as illustrated in , which illustrates a full width at half-maximum (fwhm) of ~1.8 μm.
Figure 5 Cross-sectional profiles of 4, 2, and 1 μm fluorescent beads obtained using our optofluidic cell-phone fluorescent microscope (red lines) and a benchtop fluorescent microscope using a 40× objective lens (black line) are provided. These (more ...)
And finally, the captured fluorescent flow data can be digitally analyzed on modern smart phones and tablets, on a local computer (e.g., an inexpensive laptop), or at a remote PC located in, e.g., a hospital or a clinic. Depending on the setting and the local resources, one approach can be preferred over the other.