Despite development of technologies that have rendered cervical cancer largely preventable in high- income countries, women living in low and middle-income countries (LMICs) continue to bear the brunt of cervical cancer incidence and mortality. One strategy–highly sensitive human papillomavirus (HPV) testing–has been shown to reduce the incidence and mortality from cervical cancer when coupled directly with outpatient treatment for women with HPV-positive results. The effectiveness of an HPV screen & treat strategy comes at the cost of overtreatment. To address this challenge, our group has developed a low-cost, mobile colposcope, called the Pocket Colposcope, which has shown high concordance with standard colposcopy. Further, we have developed a Colposcopy Automated Risk Evaluation (Pocket CARE) algorithm using convolutional neural networks to diagnose cervical pre-cancer among women who are screen positive, facilitating triage of patients without complex equipment or experts.

Digital video otoscope is an indispensable tool in otology that allows inspection of the external auditory canal and tympanic membrane. However, existing solutions have limitations in the diagnosis of various ear diseases and portability. Here, we propose a mobile, deep learning-assisted otoscope for low-resource settings. Our deep learning architecture was trained on clinical data to identify and classify various ear diseases. To evaluate our platform, we compared its performance with the device used in the hospital practice. Our preliminary results demonstrated high diagnostic accuracy indicating a strong potential to become a viable screening solution in low-resource, non-specialist settings.

Cervical cancer disproportionately affects low and middle income countries. Automated visual evaluation – using deep learning to analyze a digital cervix photograph – has been proposed for patient management. Image quality remains a key challenge, as it can be degraded by many types of image defects. A series of such defects were artificially added to a test set consisting of N=344 digitized cervigram images from existing studies. Replicate test sets were created for different image defects: blur, recoloring, obstructions of different colors and directions, rotations, and white Gaussian noise. The augmented images were evaluated by a classifier. The two most significant image defects were blur and Gaussian noise.

We demonstrate a rapid and cost-effective particle agglutination-based sensor that is powered by holographic imaging and deep learning. A disposable capillary-based flow device is designed to host the agglutination reaction with a material cost of <2 cents/test. A mobile and inexpensive holographic microscope captures a movie of the reaction (~3 min), which is rapidly processed by trained neural networks to automatically measure the target analyte concentration within the sample. The efficacy of this mobile sensor was demonstrated by measuring C-reactive protein (CRP) concentration in human serum samples, accurately covering the high-sensitivity range (0-10µg/mL) and very high concentrations, far exceeding 10µg/mL.

Fluorescence imaging and ‘smart probes’ are an emerging point-of-care platform for microbial detection. This article proposes a proof-of-concept frugal fluorescence imaging system for the detection of bacteria such as Pseudomonas aeruginosa, Escherichia coli and Staphylococcus aureus. In this work, we investigate the capability of a trans-illuminating fluorescence imaging system to detect bacteria using a low-cost raspberry pi single board computer and a high-quality camera. This system is capable of producing high quality fluorescence and transmission images of bacterial samples with submicron resolution. We demonstrate the capabilities of the system to produce these images in conjunction with low-cost lens-based imaging optics.

Commercial imaging systems such as mobile phones are suitable for fluorescence detection of in vivo and ex vivo tissue samples. To leverage this potential, a uniform plane of excitation light is necessary to make quantitative measurements of regions within an image. We have developed a computational model to simulate the illumination of an arbitrary number of sources. Using a pattern search algorithm, the position of these sources can be determined to generate a uniform plane of excitation light. Initial studies demonstrate that 4 fiber optic sources can be used to generate uniform illumination for biopsy samples with different geometries.

Here we introduce a portable low-noise detector for characterization of faint bioluminescent light produced in in vitro, in-vivo, and ex-vivo settings including non-transgenic animals. To demonstrate sensor’s capabilities, we present functionalized bioluminescent measurements with in-vivo transgenic and non-transgenic mice, dogs, and ex-vivo human tissue. Due to its versatility, low cost and straightforward calibration, this detector is particularly useful for animal models that are not compatible with commercial bioluminescent imagers and/or for laboratories with budgetary constraints.

The development of smartphones over the last decade has led to a growth in interest of their potential to tackle numerous point-of-care diagnostic and treatment assessment applications. Despite this interest, there has been a lackluster transition of smartphones within clinical care, where the reproducibility of measurements across devices and the inability to perform image analysis within the device has hampered development. Here, we present characterization methodologies and imaging techniques, enabled by an open-source framework, for performing quantitative imaging and analysis within the iOS smartphone environment. We explore the need for RAW pixel data within fluorescence imaging and characterize the iPhone 11 to lay the foundation for its use in scientific and point-of-care applications.

We propose a compact and low-cost lensless camera that enables snapshot full-Stokes polarization imaging. Using a polarization-encoded aperture composed of three linear polarizers and a quarter wave plate on a lensless camera, our device can capture 4 images of different linear and circular polarization intensities in a single shot, which can be used to compute full-Stokes images. We can construct an ultra-thin polarization-sensitive lensless camera using a regular image sensor and perform video-rate polarimetry for various applications. We report on the design, construction and imaging performance of our prototype device.

Due to their vast availability and rapid technological advancement, smartphones have become increasingly used as a platform for development of affordable, point-of-care diagnostic devices for low-resource settings. Simultaneous detection of multiple samples can improve the effectiveness and implementation of the point-of-care detection strategies in low-resource settings, such as a primary healthcare centers, which have limited availability of skilled personnel and consumables. To address this issue, we have developed a novel optical system that allows detection of multiple analytes at the same time in a smartphone-based spectroscopic system.

For cervical cancer screening in low-HDI countries, the WHO recommends that a diagnosis is made immediately upon cervical visualization. To address variability in provider visual interpretations, we use CNNs to classify images from a low-cost, FDA-certified, portable Pocket colposcope images positive for high-grade precancer from a triaged population. We show that the combination of white-light acetic acid and green-light image stacks improves the AUC to 0.9. Pocket CARE can be used at the community level without the need for specialized physicians or inaccessible equipment, broadening access to early detection and treatment of pre-cursor lesions before they advance to cancer.

We report a mobile device based on inline holography and deep learning to directly measure the volatility of particulate matter with high-throughput. We applied this mobile device to characterize aerosols generated by electronic cigarettes (e-cigs). Our measurements revealed a negative correlation between e-cig generated particle volatility and vegetable glycerin concentration in the e-liquid. Furthermore, the addition of other chemicals, e.g., nicotine and flavoring compounds, reduced the overall volatility of e-cig generated aerosols. The presented device can monitor the dynamic behavior of e-cig aerosols in a high-throughput manner, potentially providing important information for e-cig exposure assessment via e.g., second-hand vaping.

We introduce an advanced color fundus photography using deep learning (DL) architecture for screening glaucoma in low resource setting. The proposed DL architecture is based on a convolutional neural network and trained using clinical image data from color fundus photography and optical coherence tomography. Customized hand-held device integrated with DL model detect and quantify glaucomatous damage in fundus photograph. In validation study, our approach improves the screening capability which cannot be achieved by retinal fundus photography alone. This low-cost handy device with fast-feedback software would be very adequate tool to screen glaucoma in low resource setting.

Dengue virus (DENV) is a mosquito-borne disease that poses a public health threat to sub/tropical areas worldwide. Vaccination drives require differential diagnosis of serotype-specific DENV exposure to reduce severe dengue risks, yet state-of-the-art DENV serology relies upon short-lived serotype-specific IgM or labor intensive neutralization assays. The need for high-throughput differential diagnosis is met with our multiSero platform (Byrum et al.), a screening technique capable of detecting 48 antigen-antibody pairs simultaneously, demonstrating utility for population-wide screening. Through machine-vision techniques, we quantify and classify antibody-response signals with high sensitivity to develop automated analysis pipelines capable of diagnosing serotype-specific DENV exposure.

Near infrared (NIR) in-vivo fluorescence imaging is a powerful modality capable of interrogating biological tissue in real time, at high spatial resolution, without the need for ionizing radiation, and at depths exceeding visible light imaging modalities. Several fluorescence guided surgery (FGS) imaging systems have been developed, but their complexity and cost effectively excludes low resource settings from this technology. To help make NIR FGS available globally, we developed a fluorescence imaging augmented reality Raspberry Pi-based goggle system (FAR-Pi), open-source-hardware-inspired low cost, fully wearable, compact, and battery powered redesign of our previously described goggle-based FGS system.

Additive manufacturing is an appealing option for expanding accessibility of optical fabrication; however, most efforts have thus far focused on machines with cost, size, and time limitations. Here, we demonstrate a method of optical manufacturing using a consumer-grade stereolithographic printer. Singlet and doublet lenses, as well as dispersing prisms, were created and placed in an FDM printed housing to create a consumer-grade printed spectrometer system. Elements were post-processed in two different ways using the lithographic resin and optical quality was compared between methods. Results from the printed spectrometer system closely matched those obtained using a commercially available spectrometer system.

Smartphones have been extensively demonstrated as a platform to improve the accessibility of healthcare tools in low-resource settings. We demonstrate a novel smartphone-based optical coherence tomography system (OCT) that utilizes the native optical detection and data processing capabilities of the smartphone to perform imaging at low cost and with small system footprint. In this paper, we describe the development of an Android application for real-time display and data processing from our smartphone-integrated system.