Conceived and designed the experiments: SW-H SL MM DM DMD ZJS KC. Performed the experiments: SW-H ZJS KC ARE MM. Analyzed the data: ZJS KC SW-H MR. Contributed reagents/materials/analysis tools: DMD. Wrote the paper: ZJS SW-H AG.
The authors have declared that no competing interests exist.
In this paper we report the development of two attachments to a commercial cell phone that transform the phone's integrated lens and image sensor into a 350× microscope and visible-light spectrometer. The microscope is capable of transmission and polarized microscopy modes and is shown to have 1.5 micron resolution and a usable field-of-view of
With health care costs increasing throughout the world, there is a pressing need for reducing the cost and complexity of biomedical devices
The use of low-quality, low-cost components makes sense in the context of visual pathologic inspection. In this application, trained professionals manually examine samples to observe tissue- and cellular-level disorders, often with the aid of optical dyes. In fact, the fundamental basis of pathologic diagnosis has remained essentially unchanged for more than 100 years, following the standardization of staining procedures such as hematoxylin and eosin (H & E) for tissue sections and Wright-Giemsa staining for blood samples.
Additionally, the ability to cheaply and rapidly record diffuse reflectance spectra or fluorescence spectra also has the possibility to help with medical diagnosis. One example application is in the use of a spectrometer as a pulse oximeter, where the transmitted intensity through a finger is monitored and correlated through known absorption spectra to the concentration of oxy- and deoxy-hemoglobin. Additionally, a portable spectrometer might be used for the noninvasive detection of tumors, where it has been shown that tumors differ from surrounding healthy tissue by their increased autofluorescence and differing diffuse optical properties
In this paper we propose to take advantage of the rapid improvements in commercial CMOS sensors and microscopic optics driven by the cell-phone industry to develop two common biomedical devices, namely a microscope and spectrometer, that are available as simple and inexpensive add-ons to a commercial cell phone camera. While other researchers have demonstrated previously similar devices, our attachments to the phone are much smaller, simpler, and very low cost while still maintaining an acceptable level of performance. We demonstrate their relevance in laboratory measurements as well as discuss their applications within the field of science education.
A cell phone microscope was constructed as described in
Top panel shows the cell phone microscope achieved by adding a ball lens to the cell phone camera system. Lower panel shows the cell phone spectrometer, constructed by adding a grating and collimating tube to the cell phone camera.
Image of resolution target taken with the iPhone 2G microscope, showing the ability to clearly resolve group 9 element 2, with slight distortions at the edge of the field.
However, this resolution does not extend throughout the entire field-of-view. As can be seen in the test chart, the edges of the field have a significant defocus. This is due to our use of a simple ball lens as the magnifying element, which results in significant flat-field distortions. The lens's focal plane is described by a sphere, and the portions of the field that are in focus are those that intersect that sphere. Additionally, the system suffers from pincushion distortion, or field-dependent magnification. This can be seen clearly in
Pincushion distortion and defocus due to field curvature can clearly be seen at the edge of the field-of-view.
In this paper we use a sum-modified Laplacian based multi-focus fusion algorithm developed by Qu
Left panel: image of a Wright-Giemsa stained blood smear with the center of the field in focus. Center panel: image taken with the sample plane translated towards the phone by 2 micrometers. Right panel: Fused combination of previous two images, with fusion rule determined by the sum-modified Laplacian algorithm discussed in the text.
Additionally, we also examined the effect of different shapes and focal lengths of lenses. Some representative results of these studies applied to an unstained blood sample is shown in
Left panel: image taken with a 1 mm diameter ball lens, inset shows ball lens with penny as a reference. Right panel: image taken with a GRIN lens, inset shows size of GRIN lens with respect to a penny.
Peripheral blood smears were taken from a patient with no blood-related illness, from a patient suffering from iron deficiency anemia, and a patient suffering from sickle cell anemia. The smears were prepared and stained with a modified Wright-Giemsa stain as discussed in the
Upper row: images from a traditional microscope. Bottom row: images from a cell phone microscope. Left, blood from a normal patient. Center, blood from a patient suffering from iron deficiency anemia. Right, blood from a patient suffering from sickle cell anemia.
One of our future goals is to develop a procedure to perform a partial or complete blood count. For the purposes of this paper, we show some preliminary data where we have cropped an image taken by the iPhone 4 microscope and explored automated image analysis methods to count cells. Results from an automated count utilizing the freely available CellC program developed by Selinummi
Left panel, original image. Right panel, original image with objects identified as cells shaded in red. Counting done using CellC software.
The cell phone spectrometer was constructed as described in the
Top panel, cropped image recorded by the cell phone spectrometer pointed at a standard fluorescent light fixture. White box indicates area used to determine spectrum in lower panel. Lower panel (top), an image of the spectrum corresponding to area in the white box in the top panel. Lower panel (bottom), a comparison of the spectra of the same fluorescent light fixture taken with both the cell phone spectrometer (blue) and Ocean Optics (oo) spectrometer (red).
As one very simple biologically relevant experiment, we took a tungsten bulb and recorded its spectrum as a reference. Then, holding the distance between the camera and bulb fixed, a finger was inserted over the slit of the spectrometer, allowing one to record the transmission spectrum of approximately 1 cm of tissue. These results are shown in
Upper panel, image of spectrum corresponding to a tungsten bulb. Middle panel, image of a spectrum corresponding to a tungsten bulb with a finger placed over the slit of the spectrometer. Lower panel, spectra of both the tungsten bulb and the transmission spectrum of the finger.
Finally, we measured the fluorescence spectrum of rhodamine 6G excited with a Polychrome V light source tuned to 390 nm. The spectrum of both the light source and rhodamine fluorescence were measured by both the cell phone spectrometer and Ocean Optics spectrometer. Those results are shown in
Upper panel, image of spectrum of Polychrome V light source. Middle panel, image of rhodamine 6G spectrum. Lower panel, a comparison of the excitation and emission spectra of the Polychrome V and rhodamine 6G taken with the cell phone spectrometer and validated with the commercial Ocean Optics spectrometer (oo).
Although much important research has gone into developing very sophisticated diagnostic instruments, many important medical decisions are still based on expert opinions formed by trained professionals on the basis of data gathered via conventional devices such as microscopes, cell counters, and spectrophotometers. Replacing some of these costly and monolithic instruments with cheaper, portable devices that can achieve similar performance is an attractive option for reducing the cost and infrastructure burdens that quality health care places on society. Here we present two such devices integrated into a cell phone platform.
The first instrument, a cell phone-based microscope has been shown to have a resolution of 1.5 microns in the center of its field-of-view. Although the image quality rapidly degrades in a raw image due to the use of a single ball lens, the images can still be used to accurately diagnose a variety of blood diseases, as shown in
Furthermore, we have made an initial attempt at performing a red cell count of a blood sample imaged by the cell phone microscope. Although in this case we only report results of an algorithm locating and counting cells without regard to size or shape, the CellC algorithm or one similar could be easily used to report morphometric parameters that could enable an approximate complete blood count (CBC), discriminating cells into several blood cell classes.
The second instrument, a spectrometer consisting of a grating and collimating tube attached to the cell phone's camera, is shown to be capable of recording spectra with at least 5 nm spectral resolution, with resolution and light collection efficiency being freely traded off by the choice of the slit sizes. For the purposes of this paper we chose a slit size that gave us approximately 10 nm spectral resolution and allowed enough light throughput to easily record a fluorescence spectrum as well as a diffuse transmission spectrum. In all cases, the system compared well with the commercial Ocean Optics spectrometer in terms of spectral accuracy, with all peaks overlapping as expected after calibration.
Although our spectroscopic system may not currently have the throughput to measure more weakly fluorescing compounds, or obtain high quality diffuse reflectance or transmission spectra in the presence of low signal, these are actually not intrinsic limitations to the system. For example, with a diffuse reflectance system where source and detector are coupled to the tissue through optical fibers, a designated attachment could be designed that would obviate the need for the lossy collimation tube. Additionally, the collimation tube could be replaced by an inexpensive condenser assembly similar to those found in flashlights and car headlamps that would approximately collimate light. We are currently exploring these and other options to help improve the efficiency of the detection system.
We also note that the current state of the art of cell phone cameras are based on back-thinned CMOS sensors with 8-bit dynamic range. As scientific CMOS sensors become more ubiquitous, the cell phone will surely both drive and incorporate improvements in the detector industry and it may be that in the relatively near future commercial-grade camera sensors will approach the quality of detectors used in some less-demanding scientific applications today.
To conclude, we have presented above two devices built through adding simple and inexpensive attachments to a standard cell phone. We have demonstrated basic clinical utility of these devices through some initial experiments. We note that our choice of the iPhone as the camera for this work was driven primarily by the desire to have a camera placement that allowed easy lens attachment and sample viewing, as well as a touch screen interface to avoid motion due to button pressing. However, we do not believe the choice of phone to be crucial for this setup. We have replicated some of these experiments using other phones from different manufacturers with qualitatively similar results (not shown), indicating that the choice of phone and camera specifications are not the limiting factors in the performance of our system. These promising results will form the basis of future studies as we pursue more complex and rigorous evaluations of our devices as medical instruments.
The cell-phone microscope can work in multiple modes of operation, including polarized and transmission modes. The ability of the microscope to easily obtain simple but visually striking images points to the camera's usefulness as an educational tool. Here we present some example images taken with educational goals in mind. Polarized images of sugar crystals were taken by polarizing the incident illumination, and placing a second analyzer between the ball lens and iPhone 2G. The results are compared with a conventional polarized image of the same field-of-view using a 20× microscope and shown in
Images of an sugar crystal taken through crossed polarizers. Left panel shows the image taken with a traditional microscope, right panel shows the image taken with the cell phone microscope.
Images of several commercially prepared microscope slides featuring stained samples. Top row, commercial microscope. Bottom row, cell phone microscope. Left column, pollen grains. Right two columns, plant stems.
The portability, low cost, versatility, and network connectivity of the cell-phone microscope point to potential uses in the primary, secondary and post-secondary educational environments. At the earlier ages, teachers, and students can take images of a variety of small objects from surfaces to insects that can be shared with the class. In the upper grades and college levels, the microscope can be used in the laboratory or field for obtaining close-up images of plants, thin mineral sections, rock surfaces and more that can be easily downloaded and shared as needed. As classroom microscopy equipment becomes more outdated and limited in number, the ability to perform a variety of microscopy experiments on a cell-phone becomes ever more relevant.
The electromagnetic spectrum, especially the visible “color” portion, is a widely studied subject throughout the educational continuum appearing in science standards across the country. Spectrophotometry – the quantification of light energies generated by a lightsource, passed through a material, or reflected off a surface – appears in a variety of courses at the secondary and post-secondary levels. The cell-phone spectrometer provides a simple and inexpensive way to take a qualitative measurement of the energies of light in a given light “sample” that in current classrooms is either done with equipment costing in the thousands of dollars or is simply only talked about but never experienced. In sharing the approach with several teachers we have already witnessed the use of the cell-phone spectrometer to describe and discuss the properties of various light sources (LEDs, incandescent, fluorescent. Laser and other light sources), explain how the eye perceives color and can perceive different combinations of energies as constituting the same “color”, and measure simple fluorescence. We have tested the tool's capabilities in studying reflected light from dyes, plant materials, paints and more (data not shown) as well as for measuring the presence or absence of chlorophyll in leaves and other plant parts. Overall the tool performs at the level needed for educational use and is easy to make, use and share images collected. Additionally, analysis of images with simple and freely-available image analysis tools such as ImageJ provides educational opportunities in image processing and data analysis.
The IRB (Institutional Review Board) Administration at University of California, Davis concluded that these studies are exempt from ethics review, as we used teaching samples from the Department of Pathology without personal identifiers associated with them, and samples obtained from one healthy volunteer in the lab, who is also an author on this paper. The volunteer from whom samples were collected provided both verbal and written consent for their blood to be used for this study.
Sample hematopathology slides with known diseases were identified from the teaching collection of the Pathology Department at UC Davis Medical Center, and were prepared by standard procedures in the hematopathology laboratory at the University of California, Davis. The samples had no identifiers associated with them. Additionally, sample smear slides were prepared by collecting a single drop of blood from one healthy volunteer in the lab, who is also an author on this paper, using a finger stick. The drop of blood was placed on a glass slide and smeared using a second glass slide set on an edge and dragged across the first slide creating a wedge-shaped smear of blood. The slide is then allowed to air-dry and stained following the standard procedure for Wright-Giemsa staining using the equipment in the Pathology Department at UC Davis. First, the slide is fixed by dipping it into a methyl alcohol based fixative for 15 seconds. The slide is then immediately dipped into both a methylene blue nuclear stain for 15 seconds followed by a pink counterstain for 15 seconds.
Experiments utilized Apple-brand camera-enabled cell phones (iPhone 2G and iPhone 4G). The iPhone 2G employs a 2 megapixel CMOS sensor from Micron Technologies, Inc. with overall dimensions of 3.55×2.68 mm, and is comprised of 1600×1200 2.2 micron pixels. Each pixel is composed of a red, green, and blue-filtered sub-pixel. The camera has a single plastic biconvex lens with an effective focal length of 3.36 mm. The iPhone 4G utilizes a 5 megapixel CMOS sensor manufactured by LG Innotek. The camera has a physical size of 4.54×3.40 mm, with pixel dimensions of 2592×1944 composed of 1.75 micron pixels. The camera also features an autofocusing lens, also produced by LG Innotek.
The construction of the iPhone microscope was as simple as adding a small ball lens mounted directly on top of the phone, as shown in the upper panel of
Illumination was achieved by means of white-light LED, which in some cases was covered by a piece of matte-finished adhesive tape acting as a low-grade diffuser. The LED was placed at a distance from the sample depending on the size and brightness of the source, attempting to achieve approximately collimated illumination across the field-of-view of the microscope. Simulations were performed to validate the construction of the microscope using in-house developed ray-tracing software running on the MATLAB platform (The MathWorks, Natick, MA). Simulation results demonstrated that using collimated illumination provides the system with an effectively infinite depth of focus at the expense of allowing all imperfections in the illumination beam path to imprint themselves sharply on the image. By contrast, increasing the divergence of the illumination decreases the depth of focus and allows more flexibility regarding the cleanliness and quality of the illumination path. However, as the illumination divergence increases, the portion of the field-of-view that is in focus also decreases.
Trace of
Polarized images were acquired by placing a polarizer in front of the illumination source and an analyzer between the phone and the ball lens. Microscopic images were also taken for comparison on a commercial inverted microscope (BX-51, Olympus, Center Valley, PA) equipped with a 20× objective coupled to a commercial CCD camera (DP-71, Olympus, Center Valley, PA)
Sample holding was handled in two ways. In one scenario, where the ability to have fine control over focus adjustment was necessary (
The cell phone spectrometer is shown schematically in the lower panel of