An approach for monitoring human activities and correlated postures using an earlobe-worn wearable sensor and a deep learning algorithm is proposed. The herein-used miniaturized wearable is called TRACE and is to be mounted on an earlobe, for which smaller movements are expected compared with common locations for wearable devices. This work adopts both biological [heart rate (HR)] and physical [acceleration (ACC)] signals that are simultaneously recorded in real time. The recorded raw data are sensor-fused, adequately labeled, processed, and trained by a deep learning algorithm. Using the data, several deep learning models based on different deep learning algorithms are developed and compared. Among our experimental conditions, the best classification accuracy is achieved when the biological and physical sensor signals are fused and a long-short term memory (LSTM) algorithm is used. The proposed approach successfully classified human activities and postures with high accuracy reaching $>$ 95% for two patterns of daily activities. More specifically, the proposed method achieved 97.83%, 97.31%, and 98.19% accuracy for the deep neural network (DNN), convolutional neural network (CNN), and LSTM models, respectively.

Orthostatic disorders affect 30% of older adults and increase the risk for falls. The current diagnostic standard, the blood pressure cuff, cannot capture the rapid, multifaceted dynamics of orthostasis physiology, resulting in frequent underdiagnosis. This paper demonstrates multiparameter, real-time measurement of orthostasis using TRACE, an earlobe mounted wearable developed in our group. In prior work, we demonstrated a novel metric called orthostatic hypovolemia (OHV1), the initial loss in cephalic (head) blood volume immediately upon standing. This study significantly advances our prior work by introducing an additional 2 metrics: OHV2, the cephalic blood volume deficit after the body achieves homeostasis after standing; and postural orthostatic tachycardia (POT), the increase in heart rate. The 3 metrics were evaluated in 101 older adults who wore the TRACE device during postural transitions and reported their orthostatic symptoms. Both the OHV2 and POT metrics had significant differences between the symptomatic and asymptomatic groups, with p-values of 0.0219 and 1.2x10(-6) respectively. Furthermore, the 3 combined metrics could predict symptomatic and asymptomatic individuals with 87% sensitivity and 80% specificity. To our knowledge, this is the first report of a wearable can predict orthostatic symptoms. These metrics could aid clinicians and patients in the diagnosis and evidence-based management of orthostatic disorders including orthostatic hypotension and dysautonomia.

Engineering microfluidic devices relies on the ability to manufacture sub-100 micrometer fluidic channels. Conventional lithographic methods provide high resolution but require costly exposure tools and outsourcing of masks, which extends the turnaround time to several days. The desire to accelerate design/test cycles has motivated the rapid prototyping of microfluidic channels; however, many of these methods (e.g., laser cutters, craft cutters, fused deposition modeling) have feature sizes of several hundred microns or more. In this paper, we describe a 1-day process for fabricating sub-100 µm channels, leveraging a low-cost (USD 600) 8K digital light projection (DLP) 3D resin printer. The soft lithography process includes mold printing, post-treatment, and casting polydimethylsiloxane (PDMS) elastomer. The process can produce microchannels with 44 µm lateral resolution and 25 µm height, posts as small as 400 µm, aspect ratio up to 7, structures with varying z-height, integrated reservoirs for fluidic connections, and a built-in tray for casting. We discuss strategies to obtain reliable structures, prevent mold warpage, facilitate curing and removal of PDMS during molding, and recycle the solvents used in the process. To our knowledge, this is the first low-cost 3D printer that prints extruded structures that can mold sub-100 µm channels, providing a balance between resolution, turnaround time, and cost (~USD 5 for a 2 × 5 × 0.5 cm3 chip) that will be attractive for many microfluidics labs.

Orthostatic hypotension (OH) is a reduction in blood pressure and cerebral blood flow upon standing. Presenting in nearly 30% of older adults, OH is associated with increased fall risk and other co-morbidities. The clinical standard for monitoring OH, a blood pressure (BP) cuff, fails to capture initial orthostatic hypotension (IOH) and rapid hemodynamic changes during orthostasis. In contrast, cerebral blood flow velocity (CBFV) and continuous BP requires costly instrumentation unsuitable for home monitoring. As result OH is often underdiagnosed and poorly managed. This paper proposes a novel method for monitoring OH using TRACE, a wireless earlobe sensor that quantifies orthostatic hypovolemia (OHV), postural orthostatic tachycardia (POT), and motion during a standup test. To our knowledge, OHV is a novel biometric that may be correlated with dizziness upon standing. It is extracted using a Savitzky-Golay filter algorithm with a 99% success rate. A proof of concept study on 101 older adults show that OHV is significantly different between symptomatic and asymptomatic individuals (p-value=0.0002). TRACE provides a promising, non-invasive method to remotely monitor OH in elderly individuals, which can help reduce the risk of falls and ultimately improve patient outcomes.

Sensitive, inline detection of proteins is required for post-chromatographic analyses in proteomics, cell-based assays, and drug discovery workflows. Among the common inline methods, post-column derivatization requires chemical labels, while label-free methods are either expensive (mass spectrometry) or have limited sensitivity at small length scales (UV–Vis). This paper presents a label-free detection technique based on the concept that dissolved proteins can function as surfactants and decrease the dynamic interfacial tension (IFT) of an immiscible (water–oil) interface. Existing methods for measuring IFT, such as axisymmetric drop shape analysis (ADSA), operate in batch mode and are not suitable for continuous detection. Here we show that a microfluidic flow-focusing droplet generator operating at a frequency of > 100 Hz can track IFT changes continuously, with high temporal resolution and small detection volumes. Variations in protein concentration alter the size and shape of the drops/plugs formed, and these changes can be quantified in time using a high-speed camera and in-house image processing software. Moreover, the continuously refreshing interface alleviates issues related to surface aging. Two applications are demonstrated: (1) direct injection of a single protein into a microfluidic chip. (2) post-column detection of protein mixtures separated by high performance size exclusion chromatography (SEC HPLC). Of interest, the dynamic range of protein (bovine serum albumin, BSA) was 50 -104 μg/ml without using HPLC unit. The lowest limit of detection without HPLC unit was ~ 1 μg/ml of thyroglobulin protein in a 1 nl droplet, which equates to 1 fg of total protein. When used as a detector, the aforementioned detection method offered a sensitivity of six orders of magnitude higher than conventional UV–VIS detectors.

Antibody-based drugs have been successful in a range of therapeutic categories. However, generating monoclonal antibodies is time-consuming and expensive. A common approach is Hybridoma technology, which overcomes the short life-span of IgG-secreting plasma B cells in vitro. However, many plasma B cells are lost due to the low efficiency of hybridoma cell fusion (typically <10%). Direct single B cell screening strategies have emerged to bypass hybridoma fusion and recombinatorial display, coupled with the generation of recombinant monoclonal antibodies through mammalian expression systems. Obtaining expression systems with the required productivity, specificity and stability for clinical or commercial use requires screening millions of cells and thousands of clones. Bioelectronica’s HypercellTM platform is an emerging technology used throughout antibody discovery and cell-line development to identify and isolate single, high-antibody secreting cells from large pools (~10,000,000 cells) in short periods (ca. 48 hrs). This scalable “electrofluidic” sorting system reduces time and cost by integrating antigen-detection reagents and real-time computer vision analysis to expedite single cell sorting. In this paper antigen-specific IgG-secreting hybridoma cells are identified and sorted by their secretion rate. The cells and reagents are encapsulated in a Polydisperse Oblate Dispersion system (PODs), incubated for 1–4 hours for signal gain, and loaded into the HypercellTM device for cell sorting. Alternatively, the mixture can be analyzed without sorting to produce single cell secretion “finger print” signatures that can help identify unique expression patterns and monitor cell line stability over culturing time.

A common issue in biomicrofluidic systems is that fluidic channels may become contaminated when ampiphilic molecules adsorb to the hydrophobic channel walls. Desorption rates are often measured using optical methods, many of which require a chromophore or a fluorophore label. This paper describes label-free desorption measurements using a drop frequency sensor (DFS), a microfluidic sensor reported recently by our group. The DFS is based on a surfactant retardation effect which measures the drag of surface-active agents on microdroplets generated in a tee junction. We measure desorption curves of Tween, a small molecule surfactant, from the walls of a polydimethylsiloxane channels. Typical desorption times increase with Tween concentration, ranging from 45 to 324 seconds at Tween concentrations between 10 and 1000 ppm. The limit of detection is 10 ppm.

Flow cytometers are widely used to rapidly measure characteristics of single cells. Typical laser-based instruments provide throughputs of >10,000 events/s; however, the number of measured features is typically small and apply to the entire cell volume. Imaging flow cytometers (IFC) rely instead on 2D images of the objects, providing hundreds to millions of spatially resolved features. However, the throughput of IFCs is typically lower (several thousand events/s) due to the computational overhead of 2D image processing. Here, we demonstrate a GPU-accelerated computer vision analyzer which substantially increases computational throughput. When coupled to a 300 frame per second (fps) real-time camera, the system is limited by the camera and analyzes 1260 particles/s in a 500x700 pixel video with 4-5 particles/frame. When reading from a solid state disk, the throughput increases to 4500 fps with ~3 particles per frame, resulting in a throughput of 13,500 particles/s. The reported throughput is 2.5-4X higher than existing technologies, paving the way for ultra-high throughput IFC.

Immunoglobulin G (IgG) is an essential antibody that helps fight pathogens and protect the body from infections. Measurement of IgG is useful in assessing immunity to measles, mumps, rubella (MMR), varicella, and in the diagnosis of autoimmune hepatitis, food allergy, liver diseases etc. High performance liquid chromatography (HPLC) and enzyme-linked immunosorbent assay (ELISA) are widely employed technologies for detection of IgG. HPLC-MS systems are fast and sensitive with a detection limit of 0.4 ng/mL but are bulky and expensive. ELISA, despite having a very good sensitivity of <0.02 ng/ml, requires antibody and labels. In this study, we demonstrate label-free detection of IgG using a novel sensor concept of a stagnant cap hydrodynamic retardation detector (SHRED). SHRED detects IgG by leveraging its amphiphilic structure. The IgG adsorbs to the interface of water-in-oil droplets and retards the motion of droplets by the stagnant cap effect. The frequency shift of the droplets is measured downstream with a light scattering detector. Measured parameters such as peak area, time constant, and percentage frequency drop were all found to increase linearly for the 3 different concentrations of IgG. The detection method is comparatively simple, inexpensive, label-free and real time with a current sensitivity of 150 ng/ml.

The use of monolithic normally-closed (NC) valves within microfluidic systems is currently limited by the poor scalability of mitigation strategies developed to prevent the formation of a permanent bond between the valve ‘seat’ and elastomeric membrane during fabrication. Herein we report the highly-scalable design and characterization of a slightly-open doormat (SOD) valve that exhibits properties traditionally associated with NC valves including operability at low actuation pressures, and the capacity to produce a complete seal. In addition, these valves are comparable in their scalability and in responsiveness, and exhibit a capacity to eliminate resting channel resistance that is absent in their NC counterparts. We demonstrate the utility of this novel valve design through the design and operation of a gas-on-gas multiplexer containing 32 device valves controlled by a multiplexer containing 160 SOD valves using only 10 external solenoid valves and a single common pressure source.

Computer vision (CV) is used throughout life sciences for tissue and cell imaging; this paper presents the first demonstration of CV as a biomolecular sensor. We present an integrated sensing and fluid-actuation platform that combines CV with bead-based chemistries to perform both biomolecular and cell-based assays. By rapidly detecting the size and shape of objects at 5-50 micron length scales, the CV system can accommodate a variety of assays that traditionally require separate analytical tools and workflows, providing a greater degree of data integration and automation. This paper demonstrates 3 initial use cases: 1) Semi-quantitation of a biomarker (IgG) using a clustering assay; 2) Enumeration of white blood cells expressing CD45 surface markers using functionalized nanoparticles; and 3) Cell viability detection using dyes. All assays are performed using a single CV platform.

Wearable continuous heart rate monitors (CHRM) are widely used in fitness and health applications. The majority of existing devices are either wristwatches, which utilize photoplethysmography (PPG), or chest-straps, which rely on electrocardiography (ECG). Wristwatches are popular due to their comfortable form factor but are known to have lower accuracy than chest straps and also have significant lag, which precludes their ability to track rapid changes in heart rate. In this paper, we introduce Trace, a fully self-contained, earlobe-mounted, wireless PPG CHRM which is 10-fold smaller than previously reported devices. Placement on the earlobe, where there are no muscles or tendons, is less susceptible to noise artifacts. We demonstrate the tracking of heart rate during high intensity interval training (HIIT), where heart rate changes rapidly and sensor lag cannot be tolerated. Notable is Trace's ability to measure heart rate recovery (HRR), an advanced metric commonly used by athletes to monitor fatigue, and by physicians to assess cardiovascular risk.

A digital assay is one in which the sample is partitioned into many small containers such that each partition contains a discrete number of biological entities (0, 1, 2, 3, …). A powerful technique in the biologist’s toolkit, digital assays bring a new level of precision in quantifying nucleic acids, measuring proteins and their enzymatic activity, and probing single-cell genotypes and phenotypes. Part I of this review begins with the benefits and Poisson statistics of partitioning, including sources of error. The remainder focuses on digital PCR (dPCR) for quantification of nucleic acids. We discuss five commercial instruments that partition samples into physically isolated chambers (cdPCR) or droplet emulsions (ddPCR). We compare the strengths of dPCR (absolute quantitation, precision, and ability to detect rare or mutant targets) with those of its predecessor, quantitative real-time PCR (dynamic range, larger sample volumes, and throughput). Lastly, we describe several promising applications of dPCR, including copy number variation, quantitation of circulating tumor DNA and viral load, RNA/miRNA quantitation with reverse transcription dPCR, and library preparation for next-generation sequencing. This review is intended to give a broad perspective to scientists interested in adopting digital assays into their workflows. Part II focuses on digital protein and cell assays.

Inline detectors are routinely used in biochemical analysis, but they have tradeoffs between cost, sensitivity, and the need for labeling. We introduce a novel sensor modality, the droplet frequency sensor (DFS): a sensitive, label-free, inline detector for proteins and surface-active chemicals. The DFS uses a pressure-driven, flow-focusing microfluidic drop generator whose frequency shifts systematically as analytes pass through the device. We hypothesize that the shift is due to changes in interfacial tension and hydrodynamic drag when surface-active analytes adsorb to the liquid-liquid interface. Due to the favorable scaling of interfacial phenomena, the DFS achieves high sensitivity at small length scales. Initial experiments demonstrate a limit of detection of 12.5 femtomoles (200pg) of L-galectin, and 30 femtomoles (2ng) of bovine serum albumin.

A digital assay is one in which the sample is partitioned into many containers such that each partition contains a discrete number of biological entities (0, 1, 2, 3, . . .). A powerful technique in the biologist's toolkit, digital assays bring a new level of precision in quantifying nucleic acids, measuring proteins and their enzymatic activity, and probing single-cell genotype and phenotype. Where part I of this review focused on the fundamentals of partitioning and digital PCR, part II turns its attention to digital protein and cell assays. Digital enzyme assays measure the kinetics of single proteins with enzymatic activity. Digital enzyme-linked immunoassays (ELISAs) quantify antigenic proteins with 2 to 3 log lower detection limit than conventional ELISA, making them well suited for low-abundance biomarkers. Digital cell assays probe single-cell genotype and phenotype, including gene expression, intracellular and surface proteins, metabolic activity, cytotoxicity, and transcriptomes (scRNA-seq). These methods exploit partitioning to 1) isolate single cells or proteins, 2) detect their activity via enzymatic amplification, and 3) tag them individually by coencapsulating them with molecular barcodes. When scaled, digital assays reveal stochastic differences between proteins or cells within a population, a key to understanding biological heterogeneity. This review is intended to give a broad perspective to scientists interested in adopting digital assays into their workflows.

Emerging assays in droplet microfluidics require the measurement of parameters such as drop size, velocity, trajectory, shape deformation, fluorescence intensity, and others. While micro particle image velocimetry (μPIV) and related techniques are suitable for measuring flow using tracer particles, no tool exists for tracking droplets at the granularity of a single entity. This paper presents droplet morphometry and velocimetry (DMV), a digital video processing software for time-resolved droplet analysis. Droplets are identified through a series of image processing steps which operate on transparent, translucent, fluorescent, or opaque droplets. The steps include background image generation, background subtraction, edge detection, small object removal, morphological close and fill, and shape discrimination. A frame correlation step then links droplets spanning multiple frames via a nearest neighbor search with user-defined matching criteria. Each step can be individually tuned for maximum compatibility. For each droplet found, DMV provides a time-history of 20 different parameters, including trajectory, velocity, area, dimensions, shape deformation, orientation, nearest neighbour spacing, and pixel statistics. The data can be reported via scatter plots, histograms, and tables at the granularity of individual droplets or by statistics accrued over the population. We present several case studies from industry and academic labs, including the measurement of 1) size distributions and flow perturbations in a drop generator, 2) size distributions and mixing rates in drop splitting/merging devices, 3) efficiency of single cell encapsulation devices, 4) position tracking in electrowetting operations, 5) chemical concentrations in a serial drop dilutor, 6) drop sorting efficiency of a tensiophoresis device, 7) plug length and orientation of nonspherical plugs in a serpentine channel, and 8) high throughput tracking of >250 drops in a reinjection system. Performance metrics show that highest accuracy and precision is obtained when the video resolution is >300 pixels per drop. Analysis time increases proportionally with video resolution. The current version of the software provides throughputs of 2-30 fps, suggesting the potential for real time analysis.

Cells transmit and receive information via signalling pathways. A number of studies have revealed that information is encoded in the temporal dynamics of these pathways and has highlighted how pathway architecture can influence the propagation of signals in time and space. The functional properties of pathway architecture can also be exploited by synthetic biologists to enable precise control of cellular physiology. Here, we characterised the response of a bacterial light-responsive, two-component system to oscillating signals of varying frequencies. We found that the system acted as a low-pass filter, able to respond to low-frequency oscillations and unable to respond to high-frequency oscillations. We then demonstrate that the low-pass filtering behavior can be exploited to enable precise control of gene expression using a strategy termed pulse width modulation (PWM). PWM is a common strategy used in electronics for information encoding that converts a series of digital input signals to an analog response. We further show how the PWM strategy extends the utility of bacterial optogenetic control, allowing the fine-tuning of expression levels, programming of temporal dynamics, and control of microbial physiology via manipulation of a metabolic enzyme.

The ability to precisely choose a drop volume is highly desirable for biological assays, chemical reactions, particle manufacturing, and other applications. This paper presents the first drop generator which auto-tunes plug length to a user-defined value. It uses a real-time feedback loop which measures plug length and accordingly tunes a pressuredriven flow focusing junction. The drop generator achieves a 4-fold dynamic tuning range and <10 sec tuning time. Feedback-controlled drop generators can be useful for workflows in where volume tolerances must be maintained over long periods of time, or where multiple metered volumes must be provided by a single device.