Memory-efficient low-compute segmentation algorithms for bladder-monitoring smart ultrasound devices

Post-operative urinary retention is a medical condition where patients cannot urinate despite having a full bladder. Ultrasound imaging of the bladder is used to estimate urine volume for early diagnosis and management of urine retention. Moreover, the use of bladder ultrasound can reduce the need for an indwelling urinary catheter and the risk of catheter-associated urinary tract infection. Wearable ultrasound devices combined with machine-learning based bladder volume estimation algorithms reduce the burdens of nurses in hospital settings and improve outpatient care. However, existing algorithms are memory and computation intensive, thereby demanding the use of expensive GPUs. In this paper, we develop and validate a low-compute memory-efficient deep learning model for accurate bladder region segmentation and urine volume calculation. B-mode ultrasound bladder images of 360 patients were divided into training and validation sets; another 74 patients were used as the test dataset. Our 1-bit quantized models with 4-bits and 6-bits skip connections achieved an accuracy within [Formula: see text] and [Formula: see text], respectively, of a full precision state-of-the-art neural network (NN) without any floating-point operations and with an [Formula: see text] and [Formula: see text] reduction in memory requirements to fit under 150 kB. The means and standard deviations of the volume estimation errors, relative to estimates from ground-truth clinician annotations, were [Formula: see text] ml and [Formula: see text] ml, respectively. This lightweight NN can be easily integrated on the wearable ultrasound device for automated and continuous monitoring of urine volume. Our approach can potentially be extended to other clinical applications, such as monitoring blood pressure and fetal heart rate.

Energy-Efficient Ingestible Drug Delivery System in the Dynamic Gastrointestinal Environment

Ingestible electronics are promising platforms for on-demand health monitoring and drug delivery. However, these devices and their actuators must operate in the gastrointestinal (GI) environment, which has a pH range of 1 to 8. Drug delivery systems using electrochemical dissolution of metal films are particularly susceptible to pH changes. Optimal operation in this dynamic environment stands to transform our capacity to help patients across a range of conditions. Here we present an energy-efficient ingestible electronic electrochemical drug delivery system to support subjects through operation in this dynamic environment. The proposed system consists of a drug reservoir sealed with an electrochemically dissolvable gold membrane and an electronic subsystem. An electronic subsystem controls the rate of gold dissolution by sensing and adapting to the pH of the GI environment and provides an option for energy-efficient drug delivery, reducing energy consumption by up to 42.8 %. Integrating the electronics with electrochemical drug delivery enables the proposed system to adapt to the dynamic physiological environments which makes it suitable for drug and/or therapeutic delivery at different locations in the GI tract.

Multifunctional microelectronic fibers enable wireless modulation of gut and brain neural circuits

Progress in understanding brain-viscera interoceptive signaling is hindered by a dearth of implantable devices suitable for probing both brain and peripheral organ neurophysiology during behavior. Here we describe multifunctional neural interfaces that combine the scalability and mechanical versatility of thermally drawn polymer-based fibers with the sophistication of microelectronic chips for organs as diverse as the brain and the gut. Our approach uses meters-long continuous fibers that can integrate light sources, electrodes, thermal sensors and microfluidic channels in a miniature footprint. Paired with custom-fabricated control modules, the fibers wirelessly deliver light for optogenetics and transfer data for physiological recording. We validate this technology by modulating the mesolimbic reward pathway in the mouse brain. We then apply the fibers in the anatomically challenging intestinal lumen and demonstrate wireless control of sensory epithelial cells that guide feeding behaviors. Finally, we show that optogenetic stimulation of vagal afferents from the intestinal lumen is sufficient to evoke a reward phenotype in untethered mice.

Prospective Evaluation of the Transparent, Elastomeric, Adaptable, Long-Lasting (TEAL) Respirator

N95 filtering facepiece respirators (FFR) and surgical masks are essential in reducing airborne disease transmission, particularly during the COVID-19 pandemic. However, currently available FFR's and masks have major limitations, including masking facial features, waste, and integrity after decontamination. In a multi-institutional trial, we evaluated a transparent, elastomeric, adaptable, long-lasting (TEAL) respirator to evaluate success of qualitative fit test with user experience and biometric evaluation of temperature, respiratory rate, and fit of respirator using a novel sensor. There was a 100% successful fit test among participants, with feedback demonstrating excellent or good fit (90% of participants), breathability (77.5%), and filter exchange (95%). Biometric testing demonstrated significant differences between exhalation and inhalation pressures among a poorly fitting respirator, well-fitting respirator, and the occlusion of one filter of the respirator. We have designed and evaluated a transparent elastomeric respirator and a novel biometric feedback system that could be implemented in the hospital setting.

Energy-efficient waveform for electrical stimulation of the cochlear nerve

The cochlear implant (CI) is the most successful neural prosthesis, restoring the sensation of sound in people with severe-to-profound hearing loss by electrically stimulating the cochlear nerve. Existing CIs have an external, visible unit, and an internal, surgically-placed unit. There are significant challenges associated with the external unit, as it has limited utility and CI users often report a social stigma associated with prosthesis visibility. A fully-implantable CI (FICI) would address these issues. However, the volume constraint imposed on the FICI requires less power consumption compared to today’s CI. Because neural stimulation by CI electrodes accounts for up to 90% of power consumption, reduction in stimulation power will result directly in CI power savings. To determine an energy-efficient waveform for cochlear nerve stimulation, we used a genetic algorithm approach, incorporating a computational model of a single mammalian myelinated cochlear nerve fiber coupled to a stimulator-electrode-tissue interface. The algorithm’s prediction was tested in vivo in human CI subjects. We find that implementation of a non-rectangular biphasic neural stimulation waveform may result in up to 25% charge savings and energy savings within the comfortable range of hearing for CI users. The alternative waveform may enable future development of a FICI.

Design, Modeling, and Fabrication of Chemical Vapor Deposition Grown MoS2 Circuits with E-Mode FETs for Large-Area Electronics

Two-dimensional electronics based on single-layer (SL) MoS₂ offers significant advantages for realizing large-scale flexible systems owing to its ultrathin nature, good transport properties, and stable crystalline structure. In this work, we utilize a gate first process technology for the fabrication of highly uniform enhancement mode FETs with large mobility and excellent subthreshold swing. To enable large-scale MoS₂ circuit, we also develop Verilog-A compact models that accurately predict the performance of the fabricated MoS₂ FETs as well as a parametrized layout cell for the FET to facilitate the design and layout process using computer-aided design (CAD) tools. Using this CAD flow, we designed combinational logic gates and sequential circuits (AND, OR, NAND, NOR, XNOR, latch, edge-triggered register) as well as switched capacitor dc-dc converter, which were then fabricated using the proposed flow showing excellent performance. The fabricated integrated circuits constitute the basis of a standard cell digital library that is crucial for electronic circuit design using hardware description languages. The proposed design flow provides a platform for the co-optimization of the device fabrication technology and circuits design for future ubiquitous flexible and transparent electronics using two-dimensional materials.

Small-Area, Resistive Volatile Organic Compound (VOC) Sensors Using Metal-Polymer Hybrid Film Based on Oxidative Chemical Vapor Deposition (oCVD)

We report a novel room temperature methanol sensor comprised of gold nanoparticles covalently attached to the surface of conducting copolymer films. The copolymer films are synthesized by oxidative chemical vapor deposition (oCVD), allowing substrate-independent deposition, good polymer conductivity and stability. Two different oCVD copolymers are examined: poly(3,4-ethylenedioxythiophene-co-thiophene-3-aceticacid)[poly(EDOT-co-TAA)] and poly(3,4-ehylenedioxythiophene-co-thiophene-3-ethanol)[poly(EDOT-co-3-TE)]. Covalent attachment of gold nanoparticles to the functional groups of the oCVD films results in a hybrid system with efficient sensing response to methanol. The response of the poly(EDOT-co-TAA)/Au devices is found to be superior to that of the other copolymer, confirming the importance of the linker molecules (4-aminothiophenol) in the sensing behavior. Selectivity of the sensor to methanol over n-pentane, acetone, and toluene is demonstrated. Direct fabrication on a printed circuit board (PCB) is achieved, resulting in an improved electrical contact of the organic resistor to the metal circuitry and thus enhanced sensing properties. The simplicity and low fabrication cost of the resistive element, mild working temperature, together with its compatibility with PCB substrates pave the way for its straightforward integration into electronic devices, such as wireless sensor networks.