Supervised machine learning for the prediction of infection on admission to hospital: a prospective observational cohort study.

BACKGROUND: Infection diagnosis can be challenging, relying on clinical judgement and non-specific markers of infection. We evaluated a supervised machine learning (SML) algorithm for diagnosing bacterial infection using routinely available blood parameters on presentation to hospital. METHODS: An SML algorithm was developed to classify cases into infection versus no infection using microbiology records and six available blood parameters (C-reactive protein, white cell count, bilirubin, creatinine, ALT and alkaline phosphatase) from 160203 individuals. A cohort of patients admitted to hospital over a 6 month period had their admission blood parameters prospectively inputted into the SML algorithm. They were prospectively followed up from admission to classify those who fulfilled clinical case criteria for a community-acquired bacterial infection within 72 h of admission using a pre-determined definition. Predictive ability was assessed using receiver operating characteristics (ROC) with cut-off values for optimal sensitivity and specificity explored. RESULTS: One hundred and four individuals were included prospectively. The median (range) cohort age was 65 (21-98) years. The majority were female (56/104; 54%). Thirty-six (35%) were diagnosed with infection in the first 72 h of admission. Overall, 44/104 (42%) individuals had microbiological investigations performed. Treatment was prescribed for 33/36 (92%) of infected individuals and 4/68 (6%) of those with no identifiable bacterial infection. Mean (SD) likelihood estimates for those with and without infection were significantly different. The infection group had a likelihood of 0.80 (0.09) and the non-infection group 0.50 (0.29) (P < 0.01; 95% CI: 0.20-0.40). ROC AUC was 0.84 (95% CI: 0.76-0.91). CONCLUSIONS: An SML algorithm was able to diagnose infection in individuals presenting to hospital using routinely available blood parameters.

Glucose sensors: a review of current and emerging technology.

Glucose monitoring technology has been used in the management of diabetes for three decades. Traditional devices use enzymatic methods to measure glucose concentration and provide point sample information. More recently continuous glucose monitoring devices have become available providing more detailed data on glucose excursions. In future applications the continuous glucose sensor may become a critical component of the closed loop insulin delivery system and, as such, must be selective, rapid, predictable and acceptable for continuous patient use. Many potential sensing modalities are being pursued including optical and transdermal techniques. This review aims to summarize existing technology, the methods for assessing glucose sensing devices and provide an overview of emergent sensing modalities.

Novel pH sensing semiconductor for point-of-care detection of HIV-1 viremia.

The timely detection of viremia in HIV-infected patients receiving antiviral treatment is key to ensuring effective therapy and preventing the emergence of drug resistance. In high HIV burden settings, the cost and complexity of diagnostics limit their availability. We have developed a novel complementary metal-oxide semiconductor (CMOS) chip based, pH-mediated, point-of-care HIV-1 viral load monitoring assay that simultaneously amplifies and detects HIV-1 RNA. A novel low-buffer HIV-1 pH-LAMP (loop-mediated isothermal amplification) assay was optimised and incorporated into a pH sensitive CMOS chip. Screening of 991 clinical samples (164 on the chip) yielded a sensitivity of 95% (in vitro) and 88.8% (on-chip) at >1000 RNA copies/reaction across a broad spectrum of HIV-1 viral clades. Median time to detection was 20.8 minutes in samples with >1000 copies RNA. The sensitivity, specificity and reproducibility are close to that required to produce a point-of-care device which would be of benefit in resource poor regions, and could be performed on an USB stick or similar low power device.

Selective hydrophilic modification of Parylene C films: a new approach to cell micro-patterning for synthetic biology applications.

We demonstrate a simple, accurate and versatile method to manipulate Parylene C, a material widely known for its high biocompatibility, and transform it to a substrate that can effectively control the cellular microenvironment and consequently affect the morphology and function of the cells in vitro. The Parylene C scaffolds are fabricated by selectively increasing the material's surface water affinity through lithography and oxygen plasma treatment, providing free bonds for attachment of hydrophilic biomolecules. The micro-engineered constructs were tested as culture scaffolds for rat ventricular fibroblasts and neonatal myocytes (NRVM), toward modeling the unique anisotropic architecture of native cardiac tissue. The scaffolds induced the patterning of extracellular matrix compounds and therefore of the cells, which demonstrated substantial alignment compared to typical unstructured cultures. Ca(2+) cycling properties of the NRVM measured at rates of stimulation 0.5-2 Hz were significantly modified with a shorter time to peak and time to 90% decay, and a larger fluorescence amplitude (p < 0.001). The proposed technique is compatible with standard cell culturing protocols and exhibits long-term pattern durability. Moreover, it allows the integration of monitoring modalities into the micro-engineered substrates for a comprehensive interrogation of physiological parameters.

Inductive and ultrasonic multi-tier interface for low-power, deeply implantable medical devices.

We report the development of a novel multi-tier interface which enables the wireless, noninvasive transfer of sufficient amounts of power as well as the collection and transmission of data from low-power, deeply implantable analog sensors. The interface consists of an inductive coupling subsystem and an ultrasonic subsystem. The designed and experimentally verified inductive subsystem ensures that 5 W of power is transferred across 10 mm of air gap between a single pair of PCB spiral coils with an efficiency of 83% using our prototype CMOS logic gate-based driver circuit. The implemented ultrasonic subsystem, based on ultrasonic PZT ceramic discs driven in their low-frequency, radial/planar-excitation mode, further ensures that 29 µW of power is delivered 70 mm deeper inside a homogenous liquid environment-with no acoustic matching layer employed-with an efficiency of 1%. Overall system power consumption is 2.3 W. The implant is intermittently powered every 800 msec; charging a capacitor which provides sufficient power for a duration of ~ 18 msec; sufficient for an implant µC operating at a frequency of 500 KHz to transmit a nibble (4 bits) of digitized sensed data.

Theoretical investigation of transcranial alternating current stimulation using laminar model.

Transcranial alternating current stimulation (tACs) has been gaining an increased interest in the last few years due to its capacity to modulate non-invasively high-order cortical processes, such as decision-making, language and sensory perception. Nevertheless, the underlying mechanisms of activation of this brain stimulation technique are still poorly understood. Herein, we use a finite element modelling (FEM) technique to investigate the penetration and focality of tACs in comparison to a time invariant (DC) stimulation. We show that AC stimulations generate cerebral fields that are an order of magnitude larger in the radial direction, approximately 5 times larger in the tangential direction and more focused than DC stimulations. We argue that the basis for this effect is the reduced scalp's conductivity, which minimizes the surface shunting of the stimulating currents. The outcomes of this study may help tACs users to design better protocols and interpret experimental results.

Biomimetic model of the outer plexiform layer by incorporating memristive devices.

In this paper we present a biorealistic model for the first part of the early vision of processing by incorporating memristive nanodevices. The architecture of the proposed network is based on the organization and functioning of the outer plexiform layer (OPL) in the vertebrate retina. We demonstrate that memristive devices are indeed a valuable building block for neuromorphic architectures, as their highly nonlinear and adaptive response could be exploited for establishing ultradense networks with dynamics similar to that of their biological counterparts. We particularly show that hexagonal memristive grids can be employed for faithfully emulating the smoothing effect occurring in the OPL to enhance the dynamic range of the system. In addition, we employ a memristor-based thresholding scheme for detecting the edges of grayscale images, while the proposed system is also evaluated for its adaptation and fault tolerance capacity against different light or noise conditions as well as its distinct device yields.

Offset prediction for charge-balanced stimulus waveforms.

Functional electrical stimulation with cuff electrodes involves the controlled injection of current into an electrically excitable tissue for sensory or motor rehabilitation. Some charge injected during stimulation is 'lost' at the electrode-electrolyte interface when the charge carrier is translated from an electron to an ion in the solution. The process of charge injection through chemical reactions can reduce electrode longevity and implant biocompatibility. Conventionally, the excess charge is minimized by complex hardware solutions, which are often not appropriate for robust long-term implantable solutions. Here, we present a method of waveform design that minimizes irrecoverable charge during continuous pulsing through the use of biphasic waveforms with unequally charged phases. We developed an equivalent electrical model of the electrode-electrolyte impedance based on the electrode's surface chemistry during psuedo-bipolar stimulation conditions. Simulations with the equivalent circuit determined the uncompensated charge to be a function of stimulus parameters. In vitro stimulation experiments in saline confirmed that we could preemptively compensate for the excess charge following biphasic stimulus waveforms. As a result, there was a 92% reduction in the pre-pulse potential after a pulse train with this new waveform design when compared to stimulation with conventional biphasic waveforms.

High-frequency limit of neural stimulation with ChR2.

Optogenetic technology based on light activation of genetically targeted single component opsins such as Channelrhodopsin-2 (ChR2) has been changing the way neuroscience research is conducted. This technology is becoming increasingly important for neural engineering as well. The efficiency of neural stimulation with ChR2 drops at high frequencies, often before the natural limit of the neuron is reached. This study aims to investigate the underlying mechanisms that limit the efficiency of the stimulation at high frequencies. The study analyzes the dynamics of the spikes induced by ChR2 in comparison to control stimulations using patch clamp current injection. It shows that the stimulation dynamics is limited by two mechanisms: 1) a frequency independent reduction in the conductance-to-irradiance yield due to the ChR2 light adaptation process and 2) a frequency dependent reduction in the conductance-to-current yield due to a decrease in membrane re-polarization level between spikes that weakens the ionic driving force. The effect of the first mechanism can be minimized by using ChR2 mutants with lower irradiance threshold. In contrast the effect of the second mechanism is fundamentally limited by the rate the native ion channels re-polarize the membrane potential.

A partial-current-steering biphasic stimulation driver for vestibular prostheses.

This paper describes a novel partial-current-steering stimulation circuit for implantable vestibular prostheses. The drive hardware momentarily delivers a charge-balanced asymmetric stimulus to a dummy load before steering towards the stimulation electrodes. In this fashion, power is conserved while still gaining from the benefits of current steering. The circuit has been designed to be digitally programmable as part of an implantable vestibular prosthesis. The hardware has been implemented in AMS 0.35 mum 2P4M CMOS technology.

Energy efficient medium access protocol for wireless medical body area sensor networks.

This paper presents a novel energy-efficient MAC Protocol designed specifically for wireless body area sensor networks (WBASN) focused towards pervasive healthcare applications. Wireless body area networks consist of wireless sensor nodes attached to the human body to monitor vital signs such as body temperature, activity or heart-rate. The network adopts a master-slave architecture, where the body-worn slave node periodically sends sensor readings to a central master node. Unlike traditional peer-to-peer wireless sensor networks, the nodes in this biomedical WBASN are not deployed in an ad hoc fashion. Joining a network is centrally managed and all communications are single-hop. To reduce energy consumption, all the sensor nodes are in standby or sleep mode until the centrally assigned time slot. Once a node has joined a network, there is no possibility of collision within a cluster as all communication is initiated by the central node and is addressed uniquely to a slave node. To avoid collisions with nearby transmitters, a clear channel assessment algorithm based on standard listen-before-transmit (LBT) is used. To handle time slot overlaps, the novel concept of a wakeup fallback time is introduced. Using single-hop communication and centrally controlled sleep/wakeup times leads to significant energy reductions for this application compared to more ldquoflexiblerdquo network MAC protocols such as 802.11 or Zigbee. As duty cycle is reduced, the overall power consumption approaches the standby power. The protocol is implemented in hardware as part of the Sensiumtrade system-on-chip WBASN ASIC, in a 0.13- mum CMOS process.

A silicon pancreatic Beta cell for diabetes.

The paper will consider how silicon devices such as ion-sensitive field effect transistors can be used to model metabolic functions in biology. In a first example, a biologically inspired silicon beta cell is presented to serve as the main building block of an artificial pancreas. This is to be used for real-time glucose sensing and insulin release for diabetics. This system presents the first silicon implementation of a metabolic cell capable of exhibiting variable bursting behavior upon glucose stimulation. Based on the Hodgkin and Huxley formalism, this approach achieves dynamics similar to that of biological beta cells by using devices biased in the subthreshold regime. In addition to mimicking the physiological behavior of the beta cell, the circuit achieves good power efficiency, measured to be 4.5 muW.

A non-invasive retinal prosthesis - testing the concept.

We have developed a testing platform for a novel type of retinal prosthesis. Our system uses an array of light sources as non-contact stimulators. The platform consists of an imaging system based on a CMOS camera, PC based image processing, and a stimulation address system carried out on a Field Programmable Gated Array which addresses a matrix array of LEDs. Special optics are used to focus the light from the LED array onto light sensitized cells.