Multi-Contrast Imaging and Digital Refocusing on a Mobile Microscope with a Domed LED Array

May Multi-Contrast Imaging and Digital Refocusing on a Mobile Microscope with a Domed LED Array Zachary F. Phillips 0 1 Michael V. D'Ambrosio 0 1 Lei Tian 0 1 Jared J. Rulison 0 1 Hurshal S. Patel 0 1 Nitin Sadras 0 1 Aditya V. Gande 0 1 Neil A. Switz 0 1 Daniel A. Fletcher 0 1 Laura Waller 0 1 0 1 Graduate Group in Applied Science and Technology, University of California , Berkeley, CA 94720 , USA , 2 Department of Electrical Engineering and Computer Sciences, University of California , Berkeley, CA 94720 , USA , 3 Department of Bioengineering, University of California , Berkeley, CA 94720 , USA 1 Academic Editor: Jonathan A Coles, Glasgow University , UNITED KINGDOM We demonstrate the design and application of an add-on device for improving the diagnostic and research capabilities of CellScope-a low-cost, smartphone-based point-of-care microscope. We replace the single LED illumination of the original CellScope with a programmable domed LED array. By leveraging recent advances in computational illumination, this new device enables simultaneous multi-contrast imaging with brightfield, darkfield, and phase imaging modes. Further, we scan through illumination angles to capture lightfield datasets, which can be used to recover 3D intensity and phase images without any hardware changes. This digital refocusing procedure can be used for either 3D imaging or software-only focus correction, reducing the need for precise mechanical focusing during field experiments. All acquisition and processing is performed on the mobile phone and controlled through a smartphone application, making the computational microscope compact and portable. Using multiple samples and different objective magnifications, we demonstrate that the performance of our device is comparable to that of a commercial microscope. This unique device platform extends the field imaging capabilities of CellScope, opening up new clinical and research possibilities. - Competing Interests: The authors have declared that no competing interests exist. Optical microscopy is an important tool for disease screening and diagnosis throughout the world; however, access is often restricted to centralized hospitals due to the cost and complexity of imaging hardware. Significant resources have been devoted to developing portable and affordable compact microscopes for remote clinical applications [112]. Compact microscopes based on mobile phones, including CellScope [13, 14], have demonstrated that microscopy can be effectively performed outside of hospitals and diagnostic laboratories by minimally trained healthcare workers, that images can be transmitted for confirmation of diagnosis, and that phone-based computational analysis can be used to provide automated diagnosis. These mobile microscopes complement a host of other new devices for health monitoring on smart phones [1518]. Here, we demonstrate a new variation of the CellScope microscope which incorporates recently developed techniques of computational illumination [1921] to enable new imaging modalities, including darkfield, phase imaging and digital refocusing. Contrast in clinical microscopy is usually obtained through chemical staining or tagging to enhance specific features of a sample, requiring extensive sample preparation. Label-free contrast methods which do not require staining (e.g. darkfield, phase contrast, DIC) have not yet found widespread use as a field diagnostic tool [10], in part due to significant expense and complexity of the related optical hardware. We implement here a simple phase imaging modality, Differential Phase Contrast (DPC) [2224], which requires only two images with complementary illumination patterns. The result is qualitatively similar to differential interference contrast (DIC) imaging, but has the advantage of being quantitative, which enables measurement of cell volume and dry mass [25, 26], as well as cell confluence and tracking studies [27]. Since brightfield, darkfield and phase imaging are achieved in our system simply by switching the illumination pattern, we are able to display of all three contrast modes simultaneously by synchronizing the LED illumination patterns with the camera acquisition [21]. Image capture is quasi-realtime, with a phone-processor-limited frame rate of 0.43 Hz. Given trends in the industry toward ever-more processing power, real-time (58Hz) imaging is reasonable to expect within a couple of product cycles. Using the same hardware, our Computational CellScope also implements lightfield digital refocusing, so that a sample focus can be changed after the fact (without mechanically changing focus) and 3D image stacks can be extracted for both intensity and phase modes. Further, constant focus correction (auto-focusing) can be implemented in post-processing for long timelapse studies. The digital refocusing is achieved by sequentially illuminating the sample from each of the LEDs that lie inside the numerical aperture (NA) of the objective, then post-processing to form a stack of through-focus images of intensity [19, 28] or phase contrast [20]. For thick samples, the result also provides a 3D reconstruction of the sample, similar to limited angle tomography. The computational illumination techniques used here have been previously demonstrated in a traditional microscope using a planar LED array [1921, 29]. The purpose of the LED array is to flexibly pattern illumination angles at the sample by turning on different sets of LEDs corresponding to different illumination angles. The optimal arrangement of LEDs, however, is not planar but rather a dome shape [30], which we utilize here. The domed arrangement provides significant improvements in intensity uniformity and light throughput, since LEDs can be directionally biased and arranged at uniform radius from the sample. These benefits contribute to increased signal-to-noise ratio (SNR) in the darkfield images, allowing effective high angle illumination patterning and shorter exposure times. The flexibility and speed of the programmable LED array illuminator, as well as the lack of moving parts and low cost, make the hardware very amenable to modification as a CellScope attachment. In order for our device to be practically useful in the field, we have here enforced the requirement that all of our processing and control be performed on the smartphone, without use of a PC. Thus, the device can be field-deployable as a simple add-on to CellScope. In the following sections we detail the design and performance of the hardware and software of our new Computational CellScope device. The Computational CellScope hardware involves a custom-built domed LED illuminator attached to an inverted variant of the CellScope smartphone-based microscope platform (see Fig 1). The CellScope used here is a finite-conjugate transmission microscope coupled to an Android-based Nexus 5 smartphone (LG Electronics/Google) as described in in Skandarajah, et al. [14]. Our domed illumina (...truncated)