FilterNet: A Many-to-Many Deep Learning Architecture for Time Series Classification.

In this paper, we present and benchmark FilterNet, a flexible deep learning architecture for time series classification tasks, such as activity recognition via multichannel sensor data. It adapts popular convolutional neural network (CNN) and long short-term memory (LSTM) motifs which have excelled in activity recognition benchmarks, implementing them in a many-to-many architecture to markedly improve frame-by-frame accuracy, event segmentation accuracy, model size, and computational efficiency. We propose several model variants, evaluate them alongside other published models using the Opportunity benchmark dataset, demonstrate the effect of model ensembling and of altering key parameters, and quantify the quality of the models' segmentation of discrete events. We also offer recommendations for use and suggest potential model extensions. FilterNet advances the state of the art in all measured accuracy and speed metrics when applied to the benchmarked dataset, and it can be extensively customized for other applications.

Auditory Impairments in HIV-Infected Children.

OBJECTIVES: In a cross-sectional study of human immunodeficiency virus (HIV)-infected adults, the authors showed lower distortion product otoacoustic emissions (DPOAEs) in HIV+ individuals compared with controls as well as findings consistent with a central auditory processing deficit in HIV+ adults on antiretroviral therapy. The authors hypothesized that HIV+ children would also have a higher prevalence of abnormal central and peripheral hearing test results compared with HIV- controls. DESIGN: Pure-tone thresholds, DPOAEs, and tympanometry were performed on 244 subjects (131 HIV+ and 113 HIV- subjects). Thirty-five of the HIV+, and 3 of the HIV- subjects had a history of tuberculosis treatment. Gap detection results were available for 18 HIV- and 44 HIV+ children. Auditory brainstem response results were available for 72 HIV- and 72 HIV+ children. Data from ears with abnormal tympanograms were excluded. RESULTS: HIV+ subjects were significantly more likely to have abnormal tympanograms, histories of ear drainage, tuberculosis, or dizziness. All audiometric results were compared between groups using a two-way ANOVA with HIV status and ear drainage history as grouping variables. Mean audiometric thresholds, gap detection thresholds, and auditory brainstem response latencies did not differ between groups, although the HIV+ group had a higher proportion of individuals with a hearing loss >25 dB HL in the better ear. The HIV+ group had reduced DPOAE levels (p < 0.05) at multiple frequencies compared with HIV- subjects. No relationships were found between treatment regimens or delay in starting treatment and audiological parameters. CONCLUSIONS: As expected, children with HIV+ were more likely to have a history of ear drainage, and to have abnormal tympanograms. Similar to the adult findings, the HIV+ group did not show significantly reduced audiometric thresholds, but did have significantly lower DPOAE magnitudes. These data suggest that (1) HIV+ children often have middle ear damage which complicates understanding the direct effects of HIV on the hearing system, and (2) even when corrected for confounders DPOAEs were lower in the HIV+ group. Previous studies suggest ototoxicity from antiretroviral drugs is an unlikely cause of the reduced DPOAE magnitudes. Other possibilities include effects on efferent pathways connecting to outer hair cells or a direct effect of HIV on the cochlea.

Auditory impairments in HIV-infected individuals in Tanzania.

OBJECTIVES: Abnormal hearing tests have been noted in human immunodeficiency virus (HIV)-infected patients in several studies, but the nature of the hearing deficit has not been clearly defined. The authors performed a cross-sectional study of both HIV+ and HIV- individuals in Tanzania by using an audiological test battery. The authors hypothesized that HIV+ adults would have a higher prevalence of abnormal central and peripheral hearing test results compared with HIV- controls. In addition, they anticipated that the prevalence of abnormal hearing assessments would increase with antiretroviral therapy (ART) use and treatment for tuberculosis (TB). DESIGN: Pure-tone thresholds, distortion product otoacoustic emissions (DPOAEs), tympanometry, and a gap-detection test were performed using a laptop-based hearing testing system on 751 subjects (100 HIV- in the United States, plus 651 in Dar es Salaam, Tanzania, including 449 HIV+ [130 ART- and 319 ART+], and 202 HIV-, subjects. No U.S. subjects had a history of TB treatment. In Tanzania, 204 of the HIV+ and 23 of the HIV- subjects had a history of TB treatment. Subjects completed a video and audio questionnaire about their hearing, as well as a health history questionnaire. RESULTS: HIV+ subjects had reduced DPOAE levels compared with HIV- subjects, but their hearing thresholds, tympanometry results, and gap-detection thresholds were similar. Within the HIV+ group, those on ART reported significantly greater difficulties understanding speech in noise, and were significantly more likely to report that they had difficulty understanding speech than the ART- group. The ART+ group had a significantly higher mean gap-detection threshold compared with the ART- group. No effects of TB treatment were seen. CONCLUSIONS: The fact that the ART+/ART- groups did not differ in measures of peripheral hearing ability (DPOAEs, thresholds), or middle ear measures (tympanometry), but that the ART+ group had significantly more trouble understanding speech and had higher gap-detection thresholds indicates a central processing deficit. These data suggest that: (1) hearing deficits in HIV+ individuals could be a CNS side effect of HIV infection, (2) certain ART regimens might produce CNS side effects that manifest themselves as hearing difficulties, and/or (3) some ART regimens may treat CNS HIV inadequately, perhaps due to insufficient CNS drug levels, which is reflected as a central hearing deficit. Monitoring of central hearing parameters could be used to track central effects of either HIV or ART.

Effect of PVP on the electroosmotic mobility of wet-etched glass microchannels.

We present an experimental study on the effect of polymer PVP on EOF mobility of microchannels wet etched into optical white soda lime glass, also known as Crown glass. We performed experiments to evaluate the effect of PVP concentration and pH on EOF mobility. We used on-chip capillary zone electrophoresis and a neutral fluorescent dye as a passive marker to quantify the electroosmotic flow. We performed experiments under controlled conditions by varying pH from 5.2 and 10.3 and concentration of PVP from 0 to 2.0% w/w at constant ionic strength (30 mM). Our experiments show that PVP at concentrations of 1.0% or above very effectively suppress EOF at low pH (6.6). At high pH of 10.3, PVP has a much weaker suppressing effect on EOF and increasing its concentration above about 0.5% showed negligible effect on EOF mobility. Finally, we briefly discuss the effects of pH on using PVP as an adsorbed coating. Our experiments provide useful guidelines on choosing correct pH and concentration of PVP for effective EOF suppression in glass channels.

Coupled isotachophoretic preconcentration and electrophoretic separation using bidirectional isotachophoresis.

We present a novel technique for coupling isotachophoretic preconcentration and electrophoretic separation using bidirectional isotachophoresis (ITP). Bidirectional ITP simultaneously sets up sharp ITP interfaces between relatively high- and low-mobility cations and high- and low-mobility anions. These two interfaces can migrate toward each other and be described as ion concentration shock waves. We here demonstrate a bidirectional ITP process in which we use the interaction of these anionic and cationic ITP shock waves to trigger a transformation from ITP preconcentration to electrophoretic separation. We use anionic ITP to focus anionic sample species prior to shock interaction. The interaction of the counter-propagating anionic and cationic ITP shocks then changes the local pH (and ionic strength) of the focused analyte zones. Under this new condition, the analytes no longer focus and begin to separate electrophoretically. The method provides faster and much less dispersive transition from ITP preconcentration to electrophoretic separation compared with traditional (unidirectional) transient ITP. It eliminates the need for intermediate steps between focusing and separation, such as manual buffer exchanges. We illustrate the technique with numerical simulations of species transport equations. We have validated our simulations with experimental visualization of bidirectional ITP zones. We then show the effectiveness of the technique by coupling ITP preconcentration and high-resolution separation of a 1 kbp DNA ladder via shock interaction in bidirectional ITP.

Electrophoretic mobility measurements of fluorescent dyes using on-chip capillary electrophoresis.

We present an experimental study of the effect of pH, ionic strength, and concentrations of the electroosmotic flow (EOF)-suppressing polymer polyvinylpyrrolidone (PVP) on the electrophoretic mobilities of commonly used fluorescent dyes (fluorescein, Rhodamine 6G, and Alexa Fluor 488). We performed on-chip capillary zone electrophoresis experiments to directly quantify the effective electrophoretic mobility. We use Rhodamine B as a fluorescent neutral marker (to quantify EOF) and CCD detection. We also report relevant acid dissociation constants and analyte diffusivities based on our absolute estimate (as per Nernst-Einstein diffusion). We perform well-controlled experiments in a pH range of 3-11 and ionic strengths ranging from 30 to 90 mM. We account for the influence of ionic strength on the electrophoretic transport of sample analytes through the Onsager and Fuoss theory extended for finite radii ions to obtain the absolute mobility of the fluorophores. Lastly, we briefly explore the effect of PVP on adsorption-desorption dynamics of all three analytes, with particular attention to cationic R6G.

Deep Learning Classification of Canine Behavior Using a Single Collar-Mounted Accelerometer: Real-World Validation.

Collar-mounted canine activity monitors can use accelerometer data to estimate dog activity levels, step counts, and distance traveled. With recent advances in machine learning and embedded computing, much more nuanced and accurate behavior classification has become possible, giving these affordable consumer devices the potential to improve the efficiency and effectiveness of pet healthcare. Here, we describe a novel deep learning algorithm that classifies dog behavior at sub-second resolution using commercial pet activity monitors. We built machine learning training databases from more than 5000 videos of more than 2500 dogs and ran the algorithms in production on more than 11 million days of device data. We then surveyed project participants representing 10,550 dogs, which provided 163,110 event responses to validate real-world detection of eating and drinking behavior. The resultant algorithm displayed a sensitivity and specificity for detecting drinking behavior (0.949 and 0.999, respectively) and eating behavior (0.988, 0.983). We also demonstrated detection of licking (0.772, 0.990), petting (0.305, 0.991), rubbing (0.729, 0.996), scratching (0.870, 0.997), and sniffing (0.610, 0.968). We show that the devices' position on the collar had no measurable impact on performance. In production, users reported a true positive rate of 95.3% for eating (among 1514 users), and of 94.9% for drinking (among 1491 users). The study demonstrates the accurate detection of important health-related canine behaviors using a collar-mounted accelerometer. We trained and validated our algorithms on a large and realistic training dataset, and we assessed and confirmed accuracy in production via user validation.

Lymphocyte electrotaxis in vitro and in vivo.

Electric fields are generated in vivo in a variety of physiologic and pathologic settings, including penetrating injury to epithelial barriers. An applied electric field with strength within the physiologic range can induce directional cell migration (i.e., electrotaxis) of epithelial cells, endothelial cells, fibroblasts, and neutrophils suggesting a potential role in cell positioning during wound healing. In the present study, we investigated the ability of lymphocytes to respond to applied direct current (DC) electric fields. Using a modified Transwell assay and a simple microfluidic device, we show that human PBLs migrate toward the cathode in physiologically relevant DC electric fields. Additionally, electrical stimulation activates intracellular kinase signaling pathways shared with chemotactic stimuli. Finally, video microscopic tracing of GFP-tagged immunocytes in the skin of mouse ears reveals that motile cutaneous T cells actively migrate toward the cathode of an applied DC electric field. Lymphocyte positioning within tissues can thus be manipulated by externally applied electric fields, and may be influenced by endogenous electrical potential gradients as well.

Basic principles of electrolyte chemistry for microfluidic electrokinetics. Part I: Acid-base equilibria and pH buffers.

We review fundamental and applied acid-base equilibrium chemistry useful to microfluidic electrokinetics. We present elements of acid-base equilibrium reactions and derive rules for pH calculation for simple buffers. We also present a general formulation to calculate pH of more complex, arbitrary mixtures of electrolytes, and discuss the effects of ionic strength and temperature on pH calculation. More practically, we offer advice on buffer preparation and on buffer reporting. We also discuss "real world" buffers and likely contamination sources. In particular, we discuss the effects of atmospheric carbon dioxide on buffer systems, namely, the increase in ionic strength and acidification of typical electrokinetic device buffers. In Part II of this two-paper series, we discuss the coupling of acid-base equilibria with electrolyte dynamics and electrochemistry in typical microfluidic electrokinetic systems.

Imaging and quantification of isotachophoresis zones using nonfocusing fluorescent tracers.

We present a novel method for visualizing isotachophoresis (ITP) zones. We introduce negligibly small concentrations of a fluorophore that is not focused by isotachophoresis. This nonfocusing tracer (NFT) migrates through multiple isotachophoresis zones. As it enters each zone, the NFT concentration adapts to the local electric field in each zone. ITP zones can then be visualized with a point detector or camera. The method can be used to detect, identify, and quantify unknown analyte zones and can visualize complex and even transient electrophoresis processes. This visualization technique is particularly suited to microfluidic and laboratory-on-a-chip applications, as typical fluorescence microscopes and charge-coupled device (CCD) cameras can provide high-resolution spatiotemporal data. We present a theoretical description, a methodology for identifying analytes, and experimental validation. We also visualize and analyze a complex, transient DNA ITP preconcentration and separation.