Search

Summary Background Emergency admissions for infection often lack initial diagnostic certainty. COVID-19 has highlighted a need for novel diagnostic approaches to indicate likelihood of viral infection in a pandemic setting. We aimed to derive and validate a blood transcriptional signature to detect viral infections, including COVID-19, among adults with suspected infection who presented to the emergency department. Methods Individuals (aged ≥18 years) presenting with suspected infection to an emergency department at a major teaching hospital in the UK were prospectively recruited as part of the Bioresource in Adult Infectious Diseases (BioAID) discovery cohort. Whole-blood RNA sequencing was done on samples from participants with subsequently confirmed viral, bacterial, or no infection diagnoses. Differentially expressed host genes that met additional filtering criteria were subjected to feature selection to derive the most parsimonious discriminating signature. We validated the signature via RT-qPCR in a prospective validation cohort of participants who presented to an emergency department with undifferentiated fever, and a second case-control validation cohort of emergency department participants with PCR-positive COVID-19 or bacterial infection. We assessed signature performance by calculating the area under receiver operating characteristic curves (AUROCs), sensitivities, and specificities. Findings A three-gene transcript signature, comprising HERC6, IGF1R, and NAGK, was derived from the discovery cohort of 56 participants with bacterial infections and 27 with viral infections. In the validation cohort of 200 participants, the signature differentiated bacterial from viral infections with an AUROC of 0·976 (95% CI 0·919−1·000), sensitivity of 97·3% (85·8−99·9), and specificity of 100% (63·1−100). The AUROC for C-reactive protein (CRP) was 0·833 (0·694−0·944) and for leukocyte count was 0·938 (0·840−0·986). The signature achieved higher net benefit in decision curve analysis than either CRP or leukocyte count for discriminating viral infections from all other infections. In the second validation analysis, which included SARS-CoV-2-positive participants, the signature discriminated 35 bacterial infections from 34 SARS-CoV-2-positive COVID-19 infections with AUROC of 0·953 (0·893−0·992), sensitivity 88·6%, and specificity of 94·1%. Interpretation This novel three-gene signature discriminates viral infections, including COVID-19, from other emergency infection presentations in adults, outperforming both leukocyte count and CRP, thus potentially providing substantial clinical utility in managing acute presentations with infection. Funding National Institute for Health Research, Medical Research Council, Wellcome Trust, and EU-FP7.

The ion-sensitive field-effect transistor (ISFET) is a type of electrochemical sensor with a wide range of applications. They offer advantages of being compatible with standard CMOS technology, a miniaturised form-factor, robustness, scalability and low power requirements to name a few. There are now many architectures and design strategies to construct ISFET-based systems in CMOS, and the appropriate choice of these depends heavily on the specifications of the intended application. This tutorial aims to give a designer the knowledge needed to make the best decisions for the required specifications by providing a background in theory and an overview of design trade-offs and existing approaches. Example designs which maximise performance in particular applications are discussed and practical considerations for simulation, layout and implementation are presented.

In this paper, a low voltage bandgap reference circuit has been proposed. The introduction of a modified beta multiplier bias circuit decreased the mismatch caused by the PMOS transistors opamp con...

Real-time digital polymerase chain reaction (qdPCR) coupled with machine learning (ML) methods has shown the potential to unlock scientific breakthroughs, particularly in the field of molecular diagnostics for infectious diseases. One promising application of this emerging field explores single fluorescent channel PCR multiplex by extracting target-specific kinetic and thermodynamic information contained in amplification curves, also known as data-driven multiplexing. However, accurate target classification is compromised by the presence of undesired amplification events and not ideal reaction conditions. Therefore, here, we proposed a novel framework to identify and filter out nonspecific and low-efficient reactions from qdPCR data using outlier detection algorithms purely based on sigmoidal trends of amplification curves. As a proof-of-concept, this framework is implemented to improve the classification performance of the recently reported data-driven multiplexing method called amplification curve analysis (ACA), using available published data where the ACA is demonstrated to screen carbapenemase-producing organisms in clinical isolates. Furthermore, we developed a novel strategy, named adaptive mapping filter (AMF), to adjust the percentage of outliers removed according to the number of positive counts in qdPCR. From an overall total of 152,000 amplification events, 116,222 positive amplification reactions were evaluated before and after filtering by comparing against melting peak distribution, proving that abnormal amplification curves (outliers) are linked to shifted melting distribution or decreased PCR efficiency. The ACA was applied to assess classification performance before and after AMF, showing an improved sensitivity of 1.2% when using inliers compared to a decrement of 19.6% when using outliers (ip/i-valuelt; 0.0001), removing 53.5% of all wrong melting curves based only on the amplification shape. This work explores the correlation between the kinetics of amplification curves and the thermodynamics of melting curves, and it demonstrates that filtering out nonspecific or low-efficient reactions can significantly improve the classification accuracy for cutting-edge multiplexing methodologies.

Diabetes is a chronic metabolic disorder that affects an estimated 463 million people worldwide. Aiming to improve the treatment of people with diabetes, digital health has been widely adopted in recent years and generated a huge amount of data that could be used for further management of this chronic disease. Taking advantage of this, approaches that use artificial intelligence and specifically deep learning, an emerging type of machine learning, have been widely adopted with promising results. In this paper, we present a comprehensive review of the applications of deep learning within the field of diabetes. We conducted a systematic literature search and identified three main areas that use this approach: diagnosis of diabetes, glucose management, and diagnosis of diabetes-related complications. The search resulted in the selection of 40 original research articles, of which we have summarized the key information about the employed learning models, development process, main outcomes, and baseline methods for performance evaluation. Among the analyzed literature, it is to be noted that various deep learning techniques and frameworks have achieved state-of-the-art performance in many diabetes-related tasks by outperforming conventional machine learning approaches. Meanwhile, we identify some limitations in the current literature, such as a lack of data availability and model interpretability. The rapid developments in deep learning and the increase in available data offer the possibility to meet these challenges in the near future and allow the widespread deployment of this technology in clinical settings.

Reduction of drift in Complementary Metal Oxide-Semiconductor (CMOS) Ion-Sensitive Field-Effect Transistor (ISFET) pH sensors is demonstrated using monolayer and multilayer graphene sheets. Graphene blocks the ion penetration in the CMOS passivation layers and provides the physisorption sites needed for electrical double layer formation allowing sensing. With an in-house polymer-assisted graphene transfer (PAGT) process, monolayer and multilayer graphene sheets were manually transferred on top of the sensing membrane of CMOS ISFET sensors on a 2 by 4 mm chip. Experiments with pH buffers on five different chips were performed to extract the average performance parameters of capacitive attenuation, trapped charge, sensitivity, drift and noise. The stretched exponential function, that describes dispersion processes in amorphous solids such as silicon dioxide and silicon nitride, was modified to model the dynamic drift behaviour and analyse the effect of graphene on the performance of the sensors. The results show that on average the graphene coated ISFET sensors experience about 50% reduction in drift amplitude, up to 3 times slower surface modification and perform overall better compared to the plain unmodified devices.

ESR1 mutations are important biomarkers in metastatic breast cancer. Specifically, p.E380Q and p.Y537S mutations arise in response to hormonal therapies given to patients with hormone receptor positive (HR+) breast cancer (BC). This paper demonstrates the efficacy of an ISFET based CMOS integrated Lab-on-Chip (LoC) system, coupled with variant-specific isothermal amplification chemistries, for detection and discrimination of wild type (WT) from mutant (MT) copies of the ESR1 gene. Hormonal resistant cancers often lead to increased chances of metastatic disease which leads to high mortality rates, especially in low-income regions and areas with low healthcare coverage. Design and optimization of bespoke primers was carried out and tested on a qPCR instrument and then benchmarked versus the LoC platform. Assays for detection of p.Y537S and p.E380Q were developed and tested on the LoC platform, achieving amplification in under 25 minutes and sensitivity of down to 1000 copies of DNA per reaction for both target assays. The LoC system hereby presented, is cheaper and smaller than other standard industry equivalent technologies such as qPCR and sequencing. The LoC platform proposed, has the potential to be used at a breast cancer point-of-care testing setting, offering mutational tracking of circulating tumour DNA in liquid biopsies to assist patient stratification and metastatic monitoring.

Prostate cancer (PCa) is the second most common cause of male cancer-related death worldwide. The gold standard of treatment for advanced PCa is androgen deprivation therapy (ADT). However, eventual failure of ADT is common and leads to lethal metastatic castration resistant PCa (mCRPC). As such, the detection of relevant biomarkers in the blood for drug resistance in mCRPC patients could lead to personalized treatment options. mRNA detection is often limited by the low specificity of qPCR assays which are restricted to specialised laboratories. Here, we present a novel reversetranscription loop-mediated isothermal amplification (RT-LAMP) assay and have demonstrated its capability for sensitive detection of AR-V7 and YAP1 RNA (3×101 RNA copies per reaction). This work presents a foundation for the detection of circulating mRNA in PCa on a non-invasive Lab-on-chip (LoC) device for use at point-of-care. This technique was implemented onto a Lab-on-Chip platform integrating an array of chemical sensors (ion-sensitive field-effect transistors - ISFETs) for real-time detection of RNA. Detection of RNA presence was achieved through the translation of chemical signals into electrical readouts. Validation of this technique was conducted with rapid detection (<15 min) of extracted RNA from prostate cancer cell lines 22Rv1s and DU145s.

BackgroundNucleic acid extraction is a crucial step for molecular biology applications, being a determinant for any diagnostic test procedure. Dried blood spots (DBS) have been used for decades for serology, drug monitoring, environmental investigations, and molecular studies. Nevertheless, nucleic acid extraction from DBS remains one of the main challenges to translate them to the point-of-care (POC).MethodWe have developed a fast nucleic acid extraction (NAE) method from DBS which is electricity-free and relies on cellulose filter papers (DBSFP). The performance of NAE was assessed with loop-mediated isothermal amplification (LAMP), targeting the human reference gene beta-actin. The developed method was evaluated against FTA cards and magnetic bead-based purification, using time-to-positive (min) for comparative analysis. We optimised and validated the developed method for elution (eluted disk) and disk directly in the reaction (in-situ disk), RNA and DNA detection, and whole blood stored in anticoagulants (K2EDTA and lithium heparin). Furthermore, the compatibility of DBSFP with colourimetric detection was studied to show the transferability to the POC.ResultsThe proposed DBSFP is based on grade 3 filter paper pre-treated with 8% (v/v) igepal surfactant, 1 min washing step with PBS 1X and elution in TE 1X buffer after 5 min incubation at room temperature, enabling NAE under 7 min. Obtained results were comparable to gold standard methods across tested matrices, targets and experimental conditions, demonstrating the versatility of the methodology. Lastly, eluted disk colourimetric detection was achieved with a sample-to-result turnaround time under 35 min.ConclusionsThe developed method is a fast, electricity-free, and low-cost solution for NAE from DBSFP enabling molecular testing in virtually any POC setting.

The availability of large amounts of data from continuous glucose monitoring (CGM), together with the latest advances in deep learning techniques, have opened the door to a new paradigm of algorithm design for personalized blood glucose (BG) prediction in type 1 diabetes (T1D) with superior performance. However, there are several challenges that prevent the widespread implementation of deep learning algorithms in actual clinical settings, including unclear prediction confidence and limited training data for new T1D subjects. To this end, we propose a novel deep learning framework, Fast-adaptive and Confident Neural Network (FCNN), to meet these clinical challenges. In particular, an attention-based recurrent neural network is used to learn representations from CGM input and forward a weighted sum of hidden states to an evidential output layer, aiming to compute personalized BG predictions with theoretically supported model confidence. The model-agnostic meta-learning is employed to enable fast adaptation for a new T1D subject with limited training data. The proposed framework has been validated on three clinical datasets. In particular, for a dataset including 12 subjects with T1D, FCNN achieved a root mean square error of 18.64±2.60 mg/dL and 31.07±3.62 mg/dL for 30 and 60-minute prediction horizons, respectively, which outperformed all the considered baseline methods with significant improvements. These results indicate that FCNN is a viable and effective approach for predicting BG levels in T1D. The well-trained models can be implemented in smartphone apps to improve glycemic control by enabling proactive actions through real-time glucose alerts.

In this paper, a complete Lab-on-Chip (LoC) ion imaging platform for analysing Ion-Selective Membranes (ISM) using CMOS ISFET arrays is presented. An array of 128 × 128 ISFET pixels is employed with each pixel featuring 4 transistors to bias the ISFET to a common drain amplifier. Column-level 2-step readout circuits are designed to compensate for array offset variations in a range of up to ±1 V. The chemical signal associated with a change in ionic concentration is stored and fed back to a programmable gain instrumentation amplifier for compensation and signal amplification through a global system feedback loop. This column-parallel signal pipeline also integrates an 8-bit single slope ADC and an 8-bit R-2R DAC to quantise the processed pixel output. Designed and fabricated in the TSMC 180 nm BCD process, the System-on-Chip (SoC) operates in real time with a maximum frame rate of 1000 fps, whilst occupying a silicon area of 2.3 mm × 4.5 mm. The readout platform features a high-speed digital system to perform system-level feedback compensation with a USB 3.0 interface for data streaming. With this platform we show the first reported analysis and characterisation of ISMs using an ISFETs array through capturing real-time high-speed spatio-temporal information at a resolution of 16 μm in 1000 fps, extracting time-response and sensitivity. This work paves the way of understanding the electrochemical response of ISMs, which are widely used in various biomedical applications.

Developing multiplex PCR assays requires an extensive amount of experimental testing, the number of which exponentially increases by the number of multiplexed targets. Dedicated efforts must be devoted to the design of optimal multiplex assays for specific and sensitive identification of multiple analytes in a single well reaction. Inspired by data-driven approaches, we reinvent the way of designing and developing multiplex assays by proposing a hybrid, easy-to-use workflow, named Smart-Plexer, which couples empirical testing of singleplex assays and computer simulation of multiplexing. The Smart-Plexer leverages kinetic inter-target distances among amplification curves to generate optimal multiplex PCR primer sets for accurate multi-pathogen identification. The optimal single-channel assays, together with a novel data-driven approach, Amplification Curve Analysis (ACA), were demonstrated to be capable of classifying the presence of desired targets in a single test for seven common respiratory infection pathogens.

People with Type 1 diabetes (T1D) require regular exogenous infusion of insulin to maintain their blood glucose concentration in a therapeutically adequate target range. Although the artificial pancreas and continuous glucose monitoring have been proven to be effective in achieving closed-loop control, significant challenges still remain due to the high complexity of glucose dynamics and limitations in the technology. In this work, we propose a novel deep reinforcement learning model for single-hormone (insulin) and dual-hormone (insulin and glucagon) delivery. In particular, the delivery strategies are developed by double Q-learning with dilated recurrent neural networks. For designing and testing purposes, the FDA-accepted UVA/Padova Type 1 simulator was employed. First, we performed long-term generalized training to obtain a population model. Then, this model was personalized with a small data-set of subject-specific data. In silico results show that the single and dual-hormone delivery strategies achieve good glucose control when compared to a standard basal-bolus therapy with low-glucose insulin suspension. Specifically, in the adult cohort (n = 10), percentage time in target range 70, 180 mg/dL improved from 77.6% to 80.9% with single-hormone control, and to 85.6% with dual-hormone control. In the adolescent cohort (n = 10), percentage time in target range improved from 55.5% to $\text{65.9}\%$ with single-hormone control, and to 78.8% with dual-hormone control. In all scenarios, a significant decrease in hypoglycemia was observed. These results show that the use of deep reinforcement learning is a viable approach for closed-loop glucose control in T1D.

Potentiometry and amperometry are the two most common electrochemical sensing methods. They are conventionally performed at different times, although new applications are emerging that require their simultaneous usage in a single electrochemical cell. This paper investigates the feasibility and potential drawbacks of such a setup. We use a potentiometric and an amperometric sensor to compare their output signals when they are used individually, as well as when they are combined together with a shared reference electrode. Our results in particular show that potentiometric readings with a shared reference electrode show a high correlation of 0.9981 with conventional potentiometry. In the case of amperometric sensing, the cross correlation of the simultaneous versus individual measurement is 0.9959. Furthermore, we also demonstrate concurrent measurement for potentiometry in the presence of cell current through the design of innovative test systems. This is done through measuring both varying pH values and varying concentrations of H2O2 to showcase the operation of the circuit.

Objective: Several features of the surface electromyography (sEMG) signal are related to muscle activity and fatigue. However, the time-evolution of these features are non-stationary and vary between subjects. The aim of this study is to investigate the use of adaptive algorithms to forecast sEMG feature of the trunk muscles. Methods: Shallow models and a deep convolutional neural network (CNN) were used to simultaneously learn and forecast 5 common sEMG features in real-time to provide tailored predictions. This was investigated for: up to a 25 second horizon; for 14 different muscles in the trunk; across 13 healthy subjects; while they were performing various exercises. Results: The CNN was able to forecast 25 seconds ahead of time, with 6.88% mean absolute percentage error and 3.72% standard deviation of absolute percentage error, across all the features. Moreover, the CNN outperforms the best shallow model in terms of a figure of merit combining accuracy and precision by at least 30% for all the 5 features. Conclusion: Even though the sEMG features are non-stationary and vary between subjects, adaptive learning and forecasting, especially using CNNs, can provide accurate and precise forecasts across a range of physical activities. Significance: The proposed models provide the groundwork for a wearable device which can forecast muscle fatigue in the trunk, so as to potentially prevent low back pain. Additionally, the explicit real-time forecasting of sEMG features provides a general model which can be applied to many applications of muscle activity monitoring, which helps practitioners and physiotherapists improve therapy.

The COVID-19 pandemic is a global health emergency characterized by the high rate of transmission and ongoing increase of cases globally. Rapid point-of-care (PoC) diagnostics to detect the causative virus, SARS-CoV-2, are urgently needed to identify and isolate patients, contain its spread and guide clinical management. In this work, we report the development of a rapid PoC diagnostic test (, A CMOS-based point-of-care diagnostic platform for isothermal amplification/detection of SARS-CoV-2 RNA within 20 min coupled to a smartphone for data visualization and geolocalization is presented.

ProtonDx will provide a response to the COVID-19 pandemic by bringing nucleic-acid based molecular diagnostics to the palm of your hand. It will support the deployment of the Lacewing technology, which achieves accurate, rapid, handheld and low cost detection of SARS-CoV-2 and other respiratory infections. Results are synchronized to electronic health records and geotagged for real-time surveillance of disease progression. The device was designed for use at the point of need, in places such as pharmacies, schools and workplaces. Its unique approach combines standard semiconductor technology, advanced molecular biology and 3D printed microfluidics to match the performance of a bench-based instrument. Clinical trials are currently in progress at Imperial NHS Trust, London, UK which will lead to regulatory approvals and commercialization in the next few months.

Real-time digital PCR (qdPCR) coupled with artificial intelligence has shown the potential of unlocking scientific breakthroughs, particularly in the field of molecular diagnostics for infectious diseases. One of the most promising applications is the use of machine learning (ML) methods to enable single fluorescent channel PCR multiplex by extracting target-specific kinetic and thermodynamic information contained in amplification curves. However, the robustness of such methods can be affected by the presence of undesired amplification events and nonideal reaction conditions. Therefore, here we proposed a novel framework to filter non-specific and low efficient reactions from qdPCR data using outlier detection algorithms purely based on sigmoidal trends of amplification curves. As a proof-of-concept, this framework is implemented to improve the classification performance of the recently reported ML-based Amplification Curve Analysis (ACA), using available data from a previous publication where the ACA method was used to screen carbapenemase-producing organisms in clinical isolates. Furthermore, we developed a novel strategy, named Adaptive Mapping Filter (AMF), to consider the variability of positive counts in digital PCR. Over 152,000 amplification events were analyzed. For the positive reactions, filtered and unfiltered amplification curves were evaluated by comparing against melting peak distribution, proving that abnormalities (filtered out data) are linked to shifted melting distribution or decreased PCR efficiency. The ACA was applied to compare classification accuracies before and after AMF, showing an improved sensitivity of 1.18% for inliers and 20% for outliers (p-value < 0.0001). This work explores the correlation between kinetics of amplification curves and thermodynamics of melting curves and it demonstrates that filtering out non-specific or low efficient reactions can significantly improve the classification accuracy for cutting edge multiplexing methodologies.

Loop-mediated isothermal amplification assays are currently limited to one target per reaction in the absence of melting curve analysis, molecular probes or restriction enzyme digestion. Here, we demonstrate multiplexing of five targets in a single fluorescent channel using digital LAMP and the machine learning-based method amplification curve analysis, resulting in a classification accuracy of 91.33% on 54 186 positive amplification events.

This paper presents a standalone system based on the Raspberry Pi miniature computer capable of ion-imaging and visualisation using a CMOS chemical sensing array. The sensory part utilises a $64\times 200$ ISFET-based array biased in velocity saturation that achieves high linearity in its input-output characteristics (pH-current). The system consists of a Raspberry Pi and custom PCB board and allows high portability at a low overall cost. As a result, it constitutes a portable platform capable of real-time ion sensing and visualisation. The achieved frame rate is 10FPS maximum or limited to 5FPS with real-time visualisation of the 12,800 pixel array. Moreover, we show the simultaneous and real-time calibration and readout of the ISFET array, reducing the mismatches between pixels due to uneven trapped charge, without any impact on the performance. Demonstrations further show that the system can detect changes in pH and capture visuals of dynamic ion concentration changes in a solution. Ion imaging captured at 5FPS shows HCl diffusion in a de-ionized water solution showcasing the spatial changes in the ion concentration. This constitutes a promising approach for the realisation of ion imaging platforms and portable Point-of-Care diagnostic systems using ISFET arrays.

The advent of complementary metal-oxide-semiconductor (CMOS) based lab-on-chip (LoC) technology, which combines on-chip sample handling with integrated sensing, heralds the production of revolutionary biomedical devices capable of detecting pathogenic DNA at the point-of-care (PoC) and diagnosing infectious diseases. The ion-sensitive field effect transistor’s (ISFET) inherent sensitivity to hydrogen ions makes them ideal for pairing with DNA amplification reactions to form adept LoC diagnostic devices. This paper reviews the state-of-the-art of CMOS-based LoC devices for DNA detection and discusses electric field actuation as an opportunity for enhanced sensing capabilities. Following our previous review centred on instrumentation, we provide an overview of reported architectures in recent years which have led to portable ISFET-based devices for PoC diagnostics. We then discuss the use of dielectrophoresis (DEP) for electric field manipulation of DNA using low-power, CMOS-compatible microelectrodes. The technique enables positioning of DNA close to ISFET sensing regions to localise hydrogen ion production. This provides a signal boost to the ISFETs that lowers the limit-of-detection and time-to-result of the device, as well as providing a means of on-chip sample preparation. Major challenges include electrode spoiling reactions, hindering electrokinetic effects, and the strong dependence of trapping efficiency on applied voltage. The combination of an ISFET LoC platform with a DEP actuation system is expected to represent a landmark step for the next generation of PoC diagnostics devices.

Information about the kinetics of PCR reactions are encoded in the amplification curve. However, in digital PCR (dPCR), this information is typically neglected by collapsing each amplification curve into a binary output (positive/negative). Here, we demonstrate that the large volume of raw data obtained from realtime dPCR instruments can be exploited to perform data-driven multiplexing in a single fluorescent channel using machine learning methods, by virtue of the information in the amplification curve. This new approach, referred to as amplification curve analysis (ACA), was shown using an intercalating dye (EvaGreen), reducing the cost and complexity of the assay and enabling the use of melting curve analysis for validation. As a case study, we multiplexed 3 carbapenem-resistant genes to show the impact of this approach on global challenges such as antimicrobial resistance. In the presence of single targets, we report a classification accuracy of 99.1% (N = 16188) which represents a 19.7% increase compared to multiplexing based on the final fluorescent intensity. Considering all combinations of amplification events (including coamplifications), the accuracy was shown to be 92.9% (N = 10383). To support the analysis, we derived a formula to estimate the occurrence of co-amplification in dPCR based on multivariate Poisson statistics, and suggest reducing the digital occupancy in the case of multiple targets in the same digital panel. The ACA approach takes a step towards maximizing the capabilities of existing real-time dPCR instruments and chemistries, by extracting more information from data to enable data-driven multiplexing with high accuracy. Furthermore, we expect that combining this method with existing probe-based assays will increase multiplexing capabilities significantly. We envision that once emerging point-of-care technologies can reliably capture real-time data from isothermal chemistries, the ACA method will facilitate the implementation of dPCR outside of the lab.

In this paper, we discuss the effect of mismatch due to trapped charge on floating-gate ISFET sensors which constitutes one of its major limitations when fabricated in unmodified CMOS. Especially evident when designing ISFET arrays, mismatch due to trapped charge is significant and random and causes pixels to operate outside their target operating region. Here, we show a gradient descent algorithm that uses a Programmable-Gate in-pixel capacitor to reduce the mismatch across pixels in an iterative and unsupervised manner. Furthermore, we show that this algorithm can be reduced to a single iteration step by estimating in advance the total step of the GD since it depends on a capacitive ratio. We show measured results of both approaches on a ${64}\times {200}$ ISFET array which uses a parasitic capacitance across two metal layers in-pixel to obtain a very small PG capacitor. Both approaches are demonstrated across 4 chips and significantly reduce the mismatch spread with comparable performance.

Background Machine learning (ML) is a growing field in medicine. This narrative review describes the current body of literature on ML for clinical decision support in infectious diseases (ID). Objectives We aim to inform clinicians about the use of ML for diagnosis, classification, outcome prediction and antimicrobial management in ID. Sources References for this review were identified through searches of MEDLINE/PubMed, EMBASE, Google Scholar, biorXiv, ACM Digital Library, arXiV and IEEE Xplore Digital Library up to July 2019. Content We found 60 unique ML-clinical decision support systems (ML-CDSS) aiming to assist ID clinicians. Overall, 37 (62%) focused on bacterial infections, 10 (17%) on viral infections, nine (15%) on tuberculosis and four (7%) on any kind of infection. Among them, 20 (33%) addressed the diagnosis of infection, 18 (30%) the prediction, early detection or stratification of sepsis, 13 (22%) the prediction of treatment response, four (7%) the prediction of antibiotic resistance, three (5%) the choice of antibiotic regimen and two (3%) the choice of a combination antiretroviral therapy. The ML-CDSS were developed for intensive care units (n = 24, 40%), ID consultation (n = 15, 25%), medical or surgical wards (n = 13, 20%), emergency department (n = 4, 7%), primary care (n = 3, 5%) and antimicrobial stewardship (n = 1, 2%). Fifty-three ML-CDSS (88%) were developed using data from high-income countries and seven (12%) with data from low- and middle-income countries (LMIC). The evaluation of ML-CDSS was limited to measures of performance (e.g. sensitivity, specificity) for 57 ML-CDSS (95%) and included data in clinical practice for three (5%). Implications Considering comprehensive patient data from socioeconomically diverse healthcare settings, including primary care and LMICs, may improve the ability of ML-CDSS to suggest decisions adapted to various clinical contexts. Currents gaps identified in the evaluation of ML-CDSS must also be addressed in order to know the potential impact of such tools for clinicians and patients.

Diabetes is a chronic disease affecting 415 million people worldwide. People with type 1 diabetes mellitus (T1DM) need to self-administer insulin to maintain blood glucose (BG) levels in a normal range, which is usually a very challenging task. Developing a reliable glucose forecasting model would have a profound impact on diabetes management, since it could provide predictive glucose alarms or insulin suspension at low-glucose for hypoglycemia minimisation. Recently, deep learning has shown great potential in healthcare and medical research for diagnosis, forecasting and decision-making. In this work, we introduce a deep learning model based on a dilated recurrent neural network (DRNN) to provide 30-min forecasts of future glucose levels. Using dilation, the DRNN model gains a much larger receptive field in terms of neurons aiming at capturing long-term dependencies. A transfer learning technique is also applied to make use of the data from multiple subjects. The proposed approach outperforms existing glucose forecasting algorithms, including autoregressive models (ARX), support vector regression (SVR) and conventional neural networks for predicting glucose (NNPG) (e.g. RMSE = NNPG, 22.9 mg/dL; SVR, 21.7 mg/dL; ARX, 20.1 mg/dl; DRNN, 18.9 mg/dL on the OhioT1DM dataset). The results suggest that dilated connections can improve glucose forecasting performance efficiently.

An ultra-high frame rate and high spatial resolution ion-sensing Lab-on-Chip platform using a 128 × 128 CMOS ISFET array is presented. Current mode operation is employed to facilitate high-speed operation, with the ISFET sensors biased in the triode region to provide a linear response. Sensing pixels include a reset switch to allow in-pixel calibration for non-idealities such as offset, trapped charge and drift by periodically resetting the floating gate of the ISFET sensor. Current mode row-parallel signal processing is applied throughout the readout pipeline including auto-zeroing circuits for the removal of fixed pattern noise. The 128 readout signals are multiplexed to eight high-sample-rate on-chip current mode ADCs followed by an off-chip PCIe-based readout system on a FPGA with a latency of 0.15 s. Designed in a 0.35 $\,\boldsymbol {\mu}$ m CMOS process, the complete system-on-chip occupies an area of 2.6 × 2.2 $\text{mm}^2$ with a pixel size of 18 × 12.5 $\,\boldsymbol {\mu}$ $\text{m}^2$ and the whole system achieves a frame rate of 3000 fps which is the highest reported in the literature for ISFET arrays. The platform is demonstrated in the application of real-time ion-imaging through the high-speed visualization of sodium hydroxide (NaOH) diffusion in water at 60 fps on screen in addition to slow-motion playback of ion-dynamics recorded at 3000 fps.

Breast cancer (BC) is a common cancer in women worldwide. Despite advances in treatment, up to 30% of women eventually relapse and die of metastatic breast cancer. Liquid biopsy analysis of circulating cell-free DNA fragments in the patients’ blood can monitor clonality and evolving mutations as a surrogate for tumour biopsy. Next generation sequencing platforms and digital droplet PCR can be used to profile circulating tumour DNA from liquid biopsies; however, they are expensive and time consuming for clinical use. Here, we report a novel strategy with proof-of-concept data that supports the usage of loop-mediated isothermal amplification (LAMP) to detect PIK3CA c.3140 A > G (H1047R), a prevalent BC missense mutation that is attributed to BC tumour growth. Allele-specific primers were designed and optimized to detect the p.H1047R variant following the USS-sbLAMP method. The assay was developed with synthetic DNA templates and validated with DNA from two breast cancer cell-lines and two patient tumour tissue samples through a qPCR instrument and finally piloted on an ISFET enabled microchip. This work sets a foundation for BC mutational profiling on a Lab-on-Chip device, to help the early detection of patient relapse and to monitor efficacy of systemic therapies for personalised cancer patient management.

Control of blood glucose is essential for diabetes management. Current digital therapeutic approaches for subjects with Type 1 diabetes mellitus (T1DM) such as the artificial pancreas and insulin bolus calculators leverage machine learning techniques for predicting subcutaneous glucose for improved control. Deep learning has recently been applied in healthcare and medical research to achieve state-of-the-art results in a range of tasks including disease diagnosis, and patient state prediction among others. In this work, we present a deep learning model that is capable of forecasting glucose levels with leading accuracy for simulated patient cases (RMSE = 9.38$\pm$0.71 [mg/dL] over a 30-minute horizon, RMSE = 18.87$\pm$2.25 [mg/dL] over a 60-minute horizon) and real patient cases (RMSE = 21.07$\pm$2.35 [mg/dL] for 30-minute, RMSE = 33.27$\pm$4.79\% for 60-minute). In addition, the model provides competitive performance in providing effective prediction horizon ($PH_{eff}$) with minimal time lag both in a simulated patient dataset ($PH_{eff}$ = 29.0$\pm$0.7 for 30-min and $PH_{eff}$ = 49.8$\pm$2.9 for 60-min) and in a real patient dataset ($PH_{eff}$ = 19.3$\pm$3.1 for 30-min and $PH_{eff}$ = 29.3$\pm$9.4 for 60-min). This approach is evaluated on a dataset of 10 simulated cases generated from the UVa/Padova simulator and a clinical dataset of 10 real cases each containing glucose readings, insulin bolus, and meal (carbohydrate) data. Performance of the recurrent convolutional neural network is benchmarked against four algorithms. The proposed algorithm is implemented on an Android mobile phone, with an execution time of $6$ms on a phone compared to an execution time of $780$ms on a laptop., Comment: 10 pages, 7 figures

The unmet clinical need for accurate point-of-care (POC) diagnostic tests able to discriminate bacterial from viral infection demands a solution that can be used both within healthcare settings and in the field and that can also stem the tide of antimicrobial resistance. Our approach to solve this problem is to combine the use of Host-gene signatures with our Lab-on-a-chip (LoC) technology enabling low-cost LoC expression analysis to detect Infectious Disease.Host-gene expression signatures have been extensively study as a potential tool to be implemented in the diagnosis of infectious disease. On the other hand LoC technologies using Ion-sensitive field-effect transistor (ISFET) arrays, in conjunction with isothermal chemistries, are offering a promising alternative to conventional lab-based nucleic acid amplification instruments, owing to their portable and affordable nature. Currently, the data analysis of ISFET arrays are restricted to established methods by averaging the output of every sensor to give a single time-series. This simple approach makes unrealistic assumptions, leading to insufficient performance for applications that require accurate quantification such as RNA host transcriptomics. In order to reliably quantify host-gene expression on our LoC platform enabling the classification of bacterial and viral infection on chip, we propose a novel data-driven algorithm for extracting time-to-positive values from ISFET arrays. The algorithm proposed is based on modelling sensor drift with adaptive signal processing and clustering sensors based on their behaviour with unsupervised learning methods. Results show that the approach correctly outputs a time-to-positive for all the reactions, with a high correlation to RT-qLAMP (0.85, R2 = 0.98, p < 0.01), resulting in a classification accuracy of 100 % (CI, 95 - 100). By leveraging more advanced data processing methods for ISFET arrays, this work aims to bridge the gap between translating assays from microarray analysis (expensive lab-based discovery method) to ISFET arrays (cheap point-of-care diagnostics) providing benefits on tackling infectious disease outbreak and diagnostic testing in hard-to-reach areas of the world.

The introduction of large-scale chemical sensing systems in CMOS which integrate millions of ISFET sensors have allowed applications such as DNA sequencing and fine-pixel chemical imaging systems to be realised. Using CMOS ISFETs provides advantages of digitisation directly at the sensor as well as correcting for non-linearity in its response. However, for this to be beneficial and scale, the readout circuits need to have the minimum possible footprint and power consumption. Within this context, this paper analyses an ISFET based pH-to-time readout using an inverter in the time-domain as a level-crossing detector and presents a 32 × 32 array with in-pixel digitisation for pH sensing. The inverter-based sensing pixel, controlled by a triangular waveform, converts the pH response into a time-domain signal whilst also compensating for sensor offset and thus resulting in an increase in dynamic range. The sensor pixels interface to a 15-bit asynchronous column-wise time-to-digital converter (TDC), enabling fast asynchronous conversion whilst using minimal silicon area. Parallel outputs of 32 TDC interfaces are serialised to achieve fast data throughput. This system is implemented in a standard 0.18 $\,\mu$ m CMOS technology, with a pixel size of 26 $\mu$ m × 26 $\mu$ m and a TDC area of 26 $\mu$ m × 180 $\mu$ m. Additionally, we investigate the use of additional offset compensation by having half of the array implemented with the floating gate tied down via a well diode. Measured results demonstrate the system is able to sense reliably with an average pH sensitivity of 30 mV/pH, whilst being able to compensate for sensor offset by up to $\pm$ 7 V. A resolution of 0.013 pH is achieved and noise measurements show an integrated noise of 0.08 pH within 2–500 Hz and SFDR of 42.6 dB. The total power consumption of the system is measured to be 11.286 mW when operating at a high frame rate of 1 KFPS.