Noise Reduction in Glucose Sensors

Analyte monitoring system with interference correction for accurate glucose sensing in vivo. The system uses an analyte sensor with an additional degradation indicator to measure the degradation of the main analyte indicator. By separately measuring the degradation, the system can correct for analyte level measurement errors caused by indicator degradation due to factors like oxidation. This allows more accurate glucose sensing over time as the main analyte indicator degrades.

Source

Removing independent noise from time series data using machine learning to interpolate and denoise without requiring clean data for training. The method involves learning statistical relationships between samples using a machine learning model. When applied to denoise a time series, the model is looped through the samples and in each iteration, the model predicts the current sample using nearby samples. Since independent noise can't be predicted, it is removed from the reconstructed signal. This allows denoising without designing filters or requiring clean data.

Source

Programmable gain amplifier with noise cancellation to reduce noise and improve accuracy. The amplifier uses a noise cancellation circuit in parallel with the feedback capacitor and reset switch. During sampling, a switch closes to charge the noise cancellation capacitor. During amplification, the switch opens and the noise cancellation capacitor cancels charge noise from the feedback capacitor. This reduces noise and improves amplifier accuracy.

Source

Detecting and preventing compression events in continuous glucose monitoring (CGM) devices to improve accuracy and avoid false alarms. The technique involves using multiple analyte sensors like glucose and lactate, force sensors, and compression deflectors. When compressive forces exceed a threshold, expected values for both analytes are compared to the actual values. If the differences have an inverse correlation, it indicates compression and the glucose readings are adjusted. If no inverse correlation, it means acute compression and the glucose readings are blanked. This helps avoid false alarms from compression while still compensating for it.

Source

Calibration method for continuous glucose monitoring systems that uses Kalman filtering to estimate glucose levels from interstitial sensor readings. The method involves deconvoluting interstitial sensor currents using a sequential Kalman filter to obtain estimated blood glucose levels. Blood glucose calibration parameters are then estimated from capillary glucose measurements using a conventional Kalman filter. Initial calibration parameters are estimated offline using a cubature Kalman filter. At calibration times, patient-specific prior calibration parameters from previous days are used instead of population-level priors from historical data.

Source

Monitoring glucose levels more accurately by removing noise from the biometric data, compensating for sensor drift, and tailoring the data acquisition process to different environments. The monitoring system acquires raw glucose data from a sensor, removes noise, processes the data at lower intervals to generate smoothed biometric data, computes sensor compensation, and applies it along with a delay to get more accurate glucose readings. The system can adapt the noise removal, processing, and compensation based on factors like sensor type and user activity.

Source

A system for dynamically handling substance interference with continuous glucose monitoring (CGM) devices in fast-paced healthcare environments. The system detects administrations of substances like medications, determines if they are interferents with the CGM, calculates the interference effect based on factors like concentration and timing, and executes an appropriate response to mitigate interference. This allows more accurate glucose readings in dynamic healthcare settings where multiple medications and dosages are commonly administered.

Source

Adaptive biometric sampling rate control for continuous glucose monitoring systems to optimize sampling frequency based on glucose variability. The system analyzes indicator values like standard deviation of glucose over intervals to compare and adjust the sampling rate. If variability increases, the rate increases to capture more data, and if variability decreases, the rate decreases to reduce redundancy. This dynamic adaptation allows efficient and accurate glucose monitoring without overburdening the user or sensor.

Source

Improving accuracy of in vivo glucose monitoring using implantable glucose sensors that can detect glucose levels in the body. The method involves calculating a fixed normal sensor sensitivity and a time-varying offset based on sensor data clustering. This offset compensates for abnormal sensitivity events like early signal attenuation. Applying the fixed sensitivity and time-varying offset to the sensor signals provides more accurate glucose estimates even during sensitivity anomalies.

Source

Microneedle-based continuous glucose monitoring system that improves accuracy and stability compared to conventional systems. The system uses microneedles with a specialized sensor design that includes an interferent blocking agent, a diffusion-limiting layer, and an attachment enhancer. The interferent blocking agent fills voids in the sensor layer to prevent interferents from accessing the electrode. The diffusion-limiting layer reduces glucose diffusion to the sensor. The attachment enhancer decreases variability. This improves accuracy and stability by reducing interference currents and variability compared to conventional microneedle glucose sensors.

Source

In this study, a deep learning-based denoising autoencoder approach is proposed to increase the robustness of near-infrared spectroscopy data random noise and improve quantitative modeling accuracy. Artificial Gaussian at four different levels (10, 15, 20, 25 dB) was added spectra obtained from milk samples mimic real measurement conditions. The noisy were denoised by processing with an architecture consisting fully connected layers. removal performance quantitatively evaluated both theoretical measured signal-to-noise ratio values. results show that AE model significantly improves spectral signal quality all levels. particular, lowest level (10 dB), value nearly tripled 29.6 dB autoencoder. At other levels, average 18-20 observed in spectra. second stage Partial Least Squares Regression models built using cleaned on test set root mean square error coefficient determination. achieved lower higher determination values Especially 10 level, increased 0.44 0.71, while means decreased 0.60 0.43. can effectively reduce performance. This provides effective solution reliability accuracy anal... Read More

Source

For four-channel photoplethysmograms (PPGs), this paper employs quaternion-valued medians as features for performing non-invasive blood glucose estimation. However, the PPGs are contaminated by noise, also noise. To address issue, principal component analysis (PCA) is employed denoising. In particular, covariance matrix of computed and eigen vectors found. Then, found these represented real-valued vectors. By applying PCA to reconstructing denoised four-dimensional vectors, denoised. Next, medians. Compared traditional denoising methods feature extraction that performed in individual channels, via fusing all together. Hence, hidden relationships among four channels exploited. Finally, random forest used estimate levels (BGLs). Our proposed PCA-based compared median each channel other such time-domain frequency-domain features. Here, effectiveness robustness our method demonstrated using two datasets. The computer numerical simulation results indicate outperform existing

Source

Introduction and Objective: Continuous glucose monitoring (CGM) is a diagnostic tool widely used to monitor levels in patients with diabetes guide insulin dose adjustments outpatients. In the critical setting of inpatient care, however, special regulatory requirements on measurements necessitate accuracy CGM as prerequisite for its integration into clinical decision support. Methods: Strategies optimize were explored meet specific in-hospital care. Accuracy was assessed 226 using paired (CGM-G) point-of-care (POC-G) measurements, by calculating Mean Absolute Relative Difference (MARD) estimation numbers different zones Clarke Error Grid (CEG). Using raw data, dynamic, patient-specific algorithm developed minimize MARD through time lag optimization linear modeling. The integrated workflow applied second cohort 24 patients. Results: Data analysis showed an initial 10.18% 99.02% data points located A B CEG. application improved 4.33%. Integration reduced intrapersonal 5.58%, demonstrated significant improvements performance individual level. Conclusion: Patient-specific algorithmic refi... Read More

Source

Improving the accuracy of continuous glucose monitoring (CGM) sensors by taking into account insulin delivery information during calibration. The method involves delaying or modifying the calibration routine when insulin is being delivered, as insulin affects the glucose measurement. The delay allows the glucose level to stabilize after insulin administration. When calibration is allowed, parameters like insulin dose and delivery time are factored in to improve sensitivity determination. This leverages the CGM's capability to measure interstitial glucose, which lags blood glucose, to compensate for insulin's effect on the CGM. The goal is to provide more accurate CGM readings by accounting for insulin's impact on glucose levels during calibration.

Source

Accurately detecting failure of continuous glucose monitoring (CGM) sensors to avoid false alarms and patient disturbance while still ensuring safety. The method involves evaluating both the number of samples below a threshold and the temperature output. Sensor failure is signaled when a certain number of samples are below threshold within specific time windows, with lower temperature required for longer windows. A machine learning model can also be used to analyze sensor and temperature data.

Source

Correcting luminescent signals from implants to accurately monitor analytes like glucose in the body. The method involves transmitting excitation light through tissue to the implant and measuring the tissue's response. Then, excitation light is sent into the tissue and a separate detector measures the response. By comparing the tissue responses, corrected signal values are calculated to account for tissue scattering and absorption. This allows more accurate monitoring of implant signals like bioluminescence without being overwhelmed by tissue noise.

Source

Improving the accuracy and longevity of implantable glucose sensors in diabetes management by preventing edge lifting and peeling of the membranes. The sensors have interlocking segments near the sensing element that prevent the membranes from separating. This prevents edge lifting that could negatively affect side diffusion of glucose and reduce sensor accuracy. The membranes also have semi-permeable portions to regulate glucose diffusion for sensor sensitivity.

Source

This study addresses the critical need for effective glucose level measurement in managing diabetes mellitus (DM). DM is a serious, economically influential disease that has no cure at present, highlighting magnitude of prevention, control, and monitoring blood levels. systematically examined 79 articles from Google Scholar PubMed databases, focusing on non-invasive using photoacoustic system. After eliminating duplicates, 27 were reviewed. Glucose solution was predominantly used as primary sample. Fixed tunable lasers, especially near-infrared (NIR) highlighted due to their superior penetration accuracy measurements. Signal-purification techniques guarantee accurate detection by removing noise. The evaluation involved regression analysis machine learning integration determine levels statistically. choice sampling sites volunteers factor affecting accuracy. demonstrated meaningful progress development methods, particularly DM.

Source

Implantable glucose sensors with improved accuracy and reduced noise compared to conventional sensors. The sensors have a membrane over the sensing electrodes that contains a bioprotective domain. This domain has a base polymer with both hydrophilic and hydrophobic regions, along with zwitterionic compounds. The hydrophobic regions prevent excessive analyte penetration into the sensor, while the hydrophilic regions provide hydration. The zwitterionic compounds repel protein fouling. This membrane configuration reduces noise by blocking interferents, prevents analyte leakage, and resists fouling.

Source

A neural network-based method for denoising data with better noise reduction compared to traditional filtering techniques like Gaussian smoothing. The method involves training a neural network to convert noisy input data into cleaner output data. The network is trained by generating synthetic noisy data from the original clean data, converting it through the network, and estimating the original data based on the output. This iterative process optimizes the network to accurately denoise the input. The neural network denoising outperforms traditional filtering methods because it can learn complex noise patterns specific to the data distribution rather than applying a fixed filter.

Source

Minimizing distortion of blood glucose readings from a continuous glucose monitor (CGM) to provide more accurate and less confusing readings to users. The method involves determining if post-processing is needed based on sensor data characteristics, then performing post-processing like filtering and smoothing. This modified sensor data is displayed instead of the raw CGM readings. The post-processing is done selectively to avoid delay, by correcting zero values or excluding them. The display shows separate UI elements for each processed data point with distinct visual cues based on post-processing count.

Source

Low noise chemical sensor for detecting reactions with reduced noise and improved accuracy compared to conventional sensors. The sensor has a chemFET with an opening etched through the gate dielectric and filler material. A smaller secondary opening connects the filler material to the gate. This design reduces noise by isolating the reaction region from the gate. The secondary opening provides a path for ions to access the gate without directly contacting it. The filler material prevents ion diffusion through the etched opening. By isolating the reaction region from the gate, this sensor design reduces noise compared to conventional chemFETs where the reaction directly contacts the gate.

Source

The study "Performance Revision of a Wearable Antenna Sensor for Glucose Detection Utilizing Artificial Neural Networks" explores machine learning techniques to improve the accuracy and responsiveness wearable antenna sensor glucose detection. results demonstrate significant improvements in sensor's performance, particularly real-time monitoring. By fine-tuning neural network architecture, researchers achieved higher degree precision while minimizing false readings, enhancing measurements paving way more customized diabetes management strategies. findings suggest that integrating additional data sources, such as patient activity levels dietary habits, could further refine predictive capabilities system. Future work will focus on these advancements into compact, format, ensuring user comfort accessibility. Additionally, exploring potential remote monitoring features empower individuals managing their health proactively. also addresses issue non-linearity due diffraction effects from different layers microwave resonators. proposed designs are based low-cost, highly-sensitive identifyin... Read More

Source

Compensating glucose sensor readings for temperature to improve accuracy by accounting for the temperature dependence of the glucose sensor signal. The compensation involves detecting the temperature near the sensor and deciding whether it's within a threshold range. If so, the glucose value is left uncompensated. If the temperature exceeds the threshold, the glucose value is compensated using an ambient temperature measurement. This avoids presenting erroneous glucose values when the sensor environment is too hot or cold.

Source

Adaptive analog-to-digital conversion (ADC) circuits for sensors that optimize consumption and noise based on sensor signal characteristics. The ADCs adapt parameters like conversion time and filter bandwidth based on factors like signal strength, dynamic range, and noise. This allows customizing the ADC operation for different sensor signals to improve performance and efficiency.

Source

Reducing noise in a target signal while preserving high frequency components by estimating the noise using a separate filter window. The method involves estimating the noise using a filter with an independent window size on the high frequency part of the signal. This allows adjusting the window size based on the target signal, improving noise estimation accuracy. The estimated noise is used to calculate a filter coefficient for reducing noise in the target signal. This enables suppressing noise without impairing high frequency details.

Source

Accurately calibrating biometric signals from a continuous monitoring device like a glucose sensor in a way that compensates for instability in the sensor's performance over time. The method involves calculating calibration factors using reference biometric values, but if the sensor's calibration parameters are unstable, it corrects the calculated factor instead of using it directly. The correction is based on the difference between the calculated factor and the previous factor. This allows accurate calibration even when sensor stability is changing. The correction is further refined by considering the section of sensor usage where the reference biometric was taken, as stability varies over time.

Source

Method for accurately estimating blood glucose levels using interstitial fluid glucose measurements. It involves a sensor to measure interstitial glucose levels, a Kalman filter to estimate blood glucose based on the sensor readings, and a state transition model to account for factors like diffusion. The method uses filtering, noise estimation, and outlier detection to improve accuracy. The state transition models adapt based on glucose rate changes.

Source

Filtering continuous glucose monitor (CGM) signals using a modified Kalman filter to improve accuracy and reduce downtime compared to conventional techniques. The filter adapts the noise covariance terms based on detected artifacts in the raw CGM signal. When artifacts like pressure or zero crossings are identified, the process and measurement noise covariances are updated. This allows the filter to better handle non-analyte related noise that can impact CGM performance.

Source

Reducing blanking of sensor glucose signals in continuous glucose monitoring (CGM) systems by using multiple machine learning models trained for specific conditions. The CGM device inputs sensor data into multiple models, each trained for different data characteristics or abnormal conditions. The models generate predicted glucose values. The device combines the predictions to generate the displayed glucose value. This improves accuracy and reduces blanking compared to using a single model.

Source

Monitoring blood glucose levels using a continuous glucose sensor with improved noise reduction techniques. The sensor filtering uses a Kalman filter with adaptive noise covariance estimates based on innovation and residual signals. This allows better noise reduction without long filtering gaps. Artifact detection is also implemented to trigger noise covariance updates. The technique adapts the filter parameters in real-time based on signal characteristics to better track glucose levels with reduced error.

Source

Improving continuous glucose monitoring (CGM) systems by using multiple machine learning models trained for specific conditions to generate more reliable sensor glucose values. The models are trained separately for factors like sensor data availability, accuracy, and probabilistic reliance. During CGM, the models are applied and outputs averaged to generate the sensor glucose value. For outlier conditions, signatures in sensor data are identified and models adjusted to prioritize those associated with the signature. This allows customization of models for situations like limited data or high variability to improve accuracy.

Source

Improving the accuracy and reliability of continuous glucose monitoring (CGM) systems by using machine learning to detect and correct errors in complex sensor data in real time. The system trains a machine learning model to identify outlier measurements based on sensor data behavior signatures informed by criteria like iCGM. If an outlier is detected, the sensor data is blanked or not displayed. The system also trains a model to identify erroneous sensor use conditions based on error patterns. The model determines resolutions to correct the errors.

Source

Glucose sensing device that reduces the effect of noise in measuring glucose concentration. The device has a unique sensing method to minimize noise compared to conventional glucose sensors. The device generates a sensing pulse signal using just the current from the second sensing state when the voltage is different from the reference voltage. This pure current is used to generate the sensing pulse width, which is then compared to a reference pulse width to generate digital measurement data. Subtracting the noise current in the first state reduces noise in the final measurement.

Source

Improving accuracy of continuous glucose monitoring (CGM) devices by using short-term prediction to reduce random noise and calibration errors. The method involves substituting the current CGM reading with a predicted glucose value a short horizon ahead. This compensates for delays in the BG-to-IG kinetics and improves accuracy, especially at hypoglycemic levels. A Kalman filter is used for the prediction.

Source

Method to accurately determine blood glucose levels using a continuous glucose monitoring (CGM) system by leveraging the moving horizon estimation (MHE) technique. The method involves continuously measuring interstitial glucose levels using a sensor and applying MHE to estimate the blood glucose based on the interstitial measurements. This takes into account the time delay and diffusion between blood and interstitial glucose levels. The MHE allows adaptive estimation of noise variances and model parameters over time for improved accuracy compared to Kalman filtering or smoothing the interstitial signals.

Source

Implantable glucose sensor that reduces calibration needs compared to traditional glucose sensors. The sensor has two electrodes, one inside the enzymatic part of the membrane and one outside. By measuring both signals, it detects changes in non-glucose electroactive compounds like urea that affect sensor performance. This allows adaptive calibration based on stability instead of frequent calibrations. It also helps prevent false readings by filtering when glucose transport stability falls. The sensor design enables bifunctionality with enzyme and non-enzyme electrodes.

Source

Estimating actual glucose levels in a diabetic person when sensor noise or failure is present. The method involves using a probability analysis tool to determine the accuracy of the glucose sensor based on the measured results. The glucose results are then analyzed weighted by the sensor accuracy. A recursive filter estimates the actual glucose level using the weighted results. This allows more reliable glucose level estimation in the presence of sensor noise or failure.

Source

Reliability analysis of glucose sensor signals in closed loop glucose control systems to determine if the sensor is stable and accurate enough to trust for critical decisions like insulin dosing. The analysis involves evaluating metrics of potential sensor signal trends to detect significant changes in responsiveness over time. If the metrics exceed thresholds, it indicates sensor instability and may trigger alerts or adjustments to the closed loop system. This proactive sensor reliability monitoring helps mitigate issues like drifting, noise, and artifacts that could compromise patient safety.

Source

Continuous glucose monitoring device with improved noise filtering to enhance accuracy and reduce false alarms. The device uses a parameter estimation filter to process the glucose data. The filter estimates the filter parameters based on the signal-to-noise ratio to adaptively optimize the filtering. This allows the filter to better handle varying levels of noise in the glucose signals. It involves preprocessing the glucose data, estimating filter parameters, and applying the filter.

Source

Detecting erroneous measurements from continuous in-vivo blood analyte sensors like glucose monitors in medical applications. The method involves using a wavelet transform to analyze the sensor signals. It compares the sensor signal to time-shifted and frequency altered versions of a wavelet function to reveal unusual signal features. This detailed local analysis enables detecting signal irregularities during blood draw, clearance, and calibration segments that may indicate errors like dilution.

Source