Digital twins and artificial intelligence in metabolic disease research.

Digital twin technology is emerging as a transformative paradigm for personalized medicine in the management of chronic conditions. In this article, we explore the concept and key characteristics of a digital twin and its applications in chronic non-communicable metabolic disease management, with a focus on diabetes case studies. We cover various types of digital twin models, including mechanistic models based on ODEs, data-driven ML algorithms, and hybrid modeling strategies that combine the strengths of both approaches. We present successful case studies demonstrating the potential of digital twins in improving glucose outcomes for individuals with T1D and T2D, and discuss the benefits and challenges of translating digital twin research applications to clinical practice.

Artificial Intelligence and Machine Learning for Improving Glycemic Control in Diabetes: Best Practices, Pitfalls, and Opportunities.

OBJECTIVE: Artificial intelligence and machine learning are transforming many fields including medicine. In diabetes, robust biosensing technologies and automated insulin delivery therapies have created a substantial opportunity to improve health. While the number of manuscripts addressing the topic of applying machine learning to diabetes has grown in recent years, there has been a lack of consistency in the methods, metrics, and data used to train and evaluate these algorithms. This manuscript provides consensus guidelines for machine learning practitioners in the field of diabetes, including best practice recommended approaches and warnings about pitfalls to avoid. METHODS: Algorithmic approaches are reviewed and benefits of different algorithms are discussed including importance of clinical accuracy, explainability, interpretability, and personalization. We review the most common features used in machine learning applications in diabetes glucose control and provide an open-source library of functions for calculating features, as well as a framework for specifying data sets using data sheets. A review of current data sets available for training algorithms is provided as well as an online repository of data sources. SIGNIFICANCE: These consensus guidelines are designed to improve performance and translatability of new machine learning algorithms developed in the field of diabetes for engineers and data scientists.

Risky movement: Assessing fall risk in people with multiple sclerosis with wearable sensors and beacon-based smart-home monitoring.

BACKGROUND: People with multiple sclerosis (PwMS) fall frequently causing injury, social isolation, and decreased quality of life. Identifying locations and behaviors associated with high fall risk could help direct fall prevention interventions. Here we describe a smart-home system for assessing how mobility metrics relate to real-world fall risk in PwMS. METHODS: We performed a secondary analysis of a dataset of real-world falls collected from PwMS to identify patterns associated with increased fall risk. Thirty-four individuals were tracked over eight weeks with an inertial sensor comprising a triaxial accelerometer and time-of-flight radio transmitter, which communicated with beacons positioned throughout the home. We evaluated associations between locations in the home and movement behaviors prior to a fall compared with time periods when no falls occurred using metrics including gait initiation, time-spent-moving, movement length, and an entropy-based metric that quantifies movement complexity using transitions between rooms in the home. We also explored how fall risk may be related to the percent of times that a participant paused while walking (pauses-while-walking). RESULTS: Seventeen of the participants monitored sustained a total of 105 falls that were recorded. More falls occurred while walking (52%) than when stationary despite participants being largely sedentary, only walking 1.5±3.3% (median ± IQR) of the time that they were in their home. A total of 28% of falls occurred within one second of gait initiation. As the percentage of pauses-while-walking increased from 20 to 60%, the likelihood of a fall increased by nearly 3 times from 0.06 to 0.16%. Movement complexity, which was quantified using the entropy of room transitions, was significantly higher in the 10 min preceding falls compared with other 10-min time segments not preceding falls (1.15 ± 0.47 vs. 0.96 ± 0.24, P = 0.02). Path length was significantly longer (151.3 ± 156.1 m vs. 95.0 ± 157.2 m, P = 0.003) in the ten minutes preceding a fall compared with non-fall periods. Fall risk also varied among rooms but not consistently across participants. CONCLUSIONS: Movement metrics derived from wearable sensors and smart-home tracking systems are associated with fall risk in PwMS. More pauses-while-walking, and more complex, longer movement trajectories are associated with increased fall risk. FUNDING: Department of Veterans Affairs (RX001831-01A1). National Science Foundation (#2030859).

Combining uncertainty-aware predictive modeling and a bedtime Smart Snack intervention to prevent nocturnal hypoglycemia in people with type 1 diabetes on multiple daily injections.

OBJECTIVE: Nocturnal hypoglycemia is a known challenge for people with type 1 diabetes, especially for physically active individuals or those on multiple daily injections. We developed an evidential neural network (ENN) to predict at bedtime the probability and timing of nocturnal hypoglycemia (0-4 vs 4-8 h after bedtime) based on several glucose metrics and physical activity patterns. We utilized these predictions in silico to prescribe bedtime carbohydrates with a Smart Snack intervention specific to the predicted minimum nocturnal glucose and timing of nocturnal hypoglycemia. MATERIALS AND METHODS: We leveraged free-living datasets collected from 366 individuals from the T1DEXI Study and Glooko. Inputs to the ENN used to model nocturnal hypoglycemia were derived from demographic information, continuous glucose monitoring, and physical activity data. We assessed the accuracy of the ENN using area under the receiver operating curve, and the clinical impact of the Smart Snack intervention through simulations. RESULTS: The ENN achieved an area under the receiver operating curve of 0.80 and 0.71 to predict nocturnal hypoglycemic events during 0-4 and 4-8 h after bedtime, respectively, outperforming all evaluated baseline methods. Use of the Smart Snack intervention reduced probability of nocturnal hypoglycemia from 23.9 ± 14.1% to 14.0 ± 13.3% and duration from 7.4 ± 7.0% to 2.4 ± 3.3% in silico. DISCUSSION: Our findings indicate that the ENN-based Smart Snack intervention has the potential to significantly reduce the frequency and duration of nocturnal hypoglycemic events. CONCLUSION: A decision support system that combines prediction of minimum nocturnal glucose and proactive recommendations for bedtime carbohydrate intake might effectively prevent nocturnal hypoglycemia and reduce the burden of glycemic self-management.

Integrating metabolic expenditure information from wearable fitness sensors into an AI-augmented automated insulin delivery system: a randomised clinical trial.

BACKGROUND: Exercise can rapidly drop glucose in people with type 1 diabetes. Ubiquitous wearable fitness sensors are not integrated into automated insulin delivery (AID) systems. We hypothesised that an AID can automate insulin adjustments using real-time wearable fitness data to reduce hypoglycaemia during exercise and free-living conditions compared with an AID not automating use of fitness data. METHODS: Our study population comprised of individuals (aged 21-50 years) with type 1 diabetes from from the Harold Schnitzer Diabetes Health Center clinic at Oregon Health and Science University, OR, USA, who were enrolled into a 76 h single-centre, two-arm randomised (4-block randomisation), non-blinded crossover study to use (1) an AID that detects exercise, prompts the user, and shuts off insulin during exercise using an exercise-aware adaptive proportional derivative (exAPD) algorithm or (2) an AID that automates insulin adjustments using fitness data in real-time through an exercise-aware model predictive control (exMPC) algorithm. Both algorithms ran on iPancreas comprising commercial glucose sensors, insulin pumps, and smartwatches. Participants executed 1 week run-in on usual therapy followed by exAPD or exMPC for one 12 h primary in-clinic session involving meals, exercise, and activities of daily living, and 2 free-living out-patient days. Primary outcome was time below range (<3·9 mmol/L) during the primary in-clinic session. Secondary outcome measures included mean glucose and time in range (3·9-10 mmol/L). This trial is registered with ClinicalTrials.gov, NCT04771403. FINDINGS: Between April 13, 2021, and Oct 3, 2022, 27 participants (18 females) were enrolled into the study. There was no significant difference between exMPC (n=24) versus exAPD (n=22) in time below range (mean [SD] 1·3% [2·9] vs 2·5% [7·0]) or time in range (63·2% [23·9] vs 59·4% [23·1]) during the primary in-clinic session. In the 2 h period after start of in-clinic exercise, exMPC had significantly lower mean glucose (7·3 [1·6] vs 8·0 [1·7] mmol/L, p=0·023) and comparable time below range (1·4% [4·2] vs 4·9% [14·4]). Across the 76 h study, both algorithms achieved clinical time in range targets (71·2% [16] and 75·5% [11]) and time below range (1·0% [1·2] and 1·3% [2·2]), significantly lower than run-in period (2·4% [2·4], p=0·0004 vs exMPC; p=0·012 vs exAPD). No adverse events occurred. INTERPRETATION: AIDs can integrate exercise data from smartwatches to inform insulin dosing and limit hypoglycaemia while improving glucose outcomes. Future AID systems that integrate exercise metrics from wearable fitness sensors may help people living with type 1 diabetes exercise safely by limiting hypoglycaemia. FUNDING: JDRF Foundation and the Leona M and Harry B Helmsley Charitable Trust, National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases.

Predicting labor onset relative to the estimated date of delivery using smart ring physiological data.

The transition from pregnancy into parturition is physiologically directed by maternal, fetal and placental tissues. We hypothesize that these processes may be reflected in maternal physiological metrics. We enrolled pregnant participants in the third-trimester (n = 118) to study continuously worn smart ring devices monitoring heart rate, heart rate variability, skin temperature, sleep and physical activity from negative temperature coefficient, 3-D accelerometer and infrared photoplethysmography sensors. Weekly surveys assessed labor symptoms, pain, fatigue and mood. We estimated the association between each metric, gestational age, and the likelihood of a participant's labor beginning prior to (versus after) the clinical estimated delivery date (EDD) of 40.0 weeks with mixed effects regression. A boosted random forest was trained on the physiological metrics to predict pregnancies that naturally passed the EDD versus undergoing onset of labor prior to the EDD. Here we report that many raw sleep, activity, pain, fatigue and labor symptom metrics are correlated with gestational age. As gestational age advances, pregnant individuals have lower resting heart rate 0.357 beats/minute/week, 0.84 higher heart rate variability (milliseconds) and shorter durations of physical activity and sleep. Further, random forest predictions determine pregnancies that would pass the EDD with accuracy of 0.71 (area under the receiver operating curve). Self-reported symptoms of labor correlate with increased gestational age and not with the timing of labor (relative to EDD) or onset of spontaneous labor. The use of maternal smart ring-derived physiological data in the third-trimester may improve prediction of the natural duration of pregnancy relative to the EDD.

Enabling fully automated insulin delivery through meal detection and size estimation using Artificial Intelligence.

We present a robust insulin delivery system that includes automated meal detection and carbohydrate content estimation using machine learning for meal insulin dosing called robust artificial pancreas (RAP). We conducted a randomized, single-center crossover trial to compare postprandial glucose control in the four hours following unannounced meals using a hybrid model predictive control (MPC) algorithm and the RAP system. The RAP system includes a neural network model to automatically detect meals and deliver a recommended meal insulin dose. The meal detection algorithm has a sensitivity of 83.3%, false discovery rate of 16.6%, and mean detection time of 25.9 minutes. While there is no significant difference in incremental area under the curve of glucose, RAP significantly reduces time above range (glucose >180 mg/dL) by 10.8% (P = 0.04) and trends toward increasing time in range (70-180 mg/dL) by 9.1% compared with MPC. Time below range (glucose <70 mg/dL) is not significantly different between RAP and MPC.

Modeling risk of hypoglycemia during and following physical activity in people with type 1 diabetes using explainable mixed-effects machine learning.

BACKGROUND: Physical activity (PA) can cause increased hypoglycemia (glucose <70 mg/dL) risk in people with type 1 diabetes (T1D). We modeled the probability of hypoglycemia during and up to 24 h following PA and identified key factors associated with hypoglycemia risk. METHODS: We leveraged a free-living dataset from Tidepool comprised of glucose measurements, insulin doses, and PA data from 50 individuals with T1D (6448 sessions) for training and validating machine learning models. We also used data from the T1Dexi pilot study that contains glucose management and PA data from 20 individuals with T1D (139 session) for assessing the accuracy of the best performing model on an independent test dataset. We used mixed-effects logistic regression (MELR) and mixed-effects random forest (MERF) to model hypoglycemia risk around PA. We identified risk factors associated with hypoglycemia using odds ratio and partial dependence analysis for the MELR and MERF models, respectively. Prediction accuracy was measured using the area under the receiver operating characteristic curve (AUROC). RESULTS: The analysis identified risk factors significantly associated with hypoglycemia during and following PA in both MELR and MERF models including glucose and body exposure to insulin at the start of PA, low blood glucose index 24 h prior to PA, and PA intensity and timing. Both models showed overall hypoglycemia risk peaking 1 h after PA and again 5-10 h after PA, which is consistent with the hypoglycemia risk pattern observed in the training dataset. Time following PA impacted hypoglycemia risk differently across different PA types. Accuracy of hypoglycemia prediction using the fixed effects of the MERF model was highest when predicting hypoglycemia during the first hour following the start of PA (AUROCVALIDATION = 0.83 and AUROCTESTING = 0.86) and decreased when predicting hypoglycemia in the 24 h after PA (AUROCVALIDATION = 0.66 and AUROCTESTING = 0.68). CONCLUSION: Hypoglycemia risk after the start of PA can be modeled using mixed-effects machine learning to identify key risk factors that may be used within decision support and insulin delivery systems. We published the population-level MERF model online for others to use.

Assessment of a Decision Support System for Adults with Type 1 Diabetes on Multiple Daily Insulin Injections.

Introduction: DailyDose is a decision support system designed to provide real-time dosing advice and weekly insulin dose adjustments for adults living with type 1 diabetes using multiple daily insulin injections. Materials and Methods: Twenty-five adults were enrolled in this single-arm study. All participants used Dexcom G6 for continuous glucose monitoring, InPen for short-acting insulin doses, and Clipsulin to track long-acting insulin doses. Participants used DailyDose on an iPhone for 8 weeks. The primary endpoint was % time in range (TIR) comparing the 2-week baseline to the final 2-week period of DailyDose use. Results: There were no significant differences between TIR or other glycemic metrics between the baseline period compared to final 2-week period of DailyDose use. TIR significantly improved by 6.3% when more than half of recommendations were accepted and followed compared with 50% or fewer recommendations (95% CI 2.5%-10.1%, P = 0.001). Conclusions: Use of DailyDose did not improve glycemic outcomes compared to the baseline period. In a post hoc analysis, accepting and following recommendations from DailyDose was associated with improved TIR. Clinical Trial Registration Number: NCT04428645.

Quantifying the impact of physical activity on future glucose trends using machine learning.

Prevention of hypoglycemia (glucose <70 mg/dL) during aerobic exercise is a major challenge in type 1 diabetes. Providing predictions of glycemic changes during and following exercise can help people with type 1 diabetes avoid hypoglycemia. A unique dataset representing 320 days and 50,000 + time points of glycemic measurements was collected in adults with type 1 diabetes who participated in a 4-arm crossover study evaluating insulin-pump therapies, whereby each participant performed eight identically designed in-clinic exercise studies. We demonstrate that even under highly controlled conditions, there is considerable intra-participant and inter-participant variability in glucose outcomes during and following exercise. Participants with higher aerobic fitness exhibited significantly lower minimum glucose and steeper glucose declines during exercise. Adaptive, personalized machine learning (ML) algorithms were designed to predict exercise-related glucose changes. These algorithms achieved high accuracy in predicting the minimum glucose and hypoglycemia during and following exercise sessions, for all fitness levels.

Incorporating Glucose Variability into Glucose Forecasting Accuracy Assessment Using the New Glucose Variability Impact Index and the Prediction Consistency Index: An LSTM Case Example.

BACKGROUND: In this work, we developed glucose forecasting algorithms trained and evaluated on a large dataset of free-living people with type 1 diabetes (T1D) using closed-loop (CL) and sensor-augmented pump (SAP) therapies; and we demonstrate how glucose variability impacts accuracy. We introduce the glucose variability impact index (GVII) and the glucose prediction consistency index (GPCI) to assess the accuracy of prediction algorithms. METHODS: A long-short-term-memory (LSTM) neural network was designed to predict glucose up to 60 minutes in the future using continuous glucose measurements and insulin data collected from 175 people with T1D (41,318 days) and evaluated on 75 people (11,333 days) from the Tidepool Big Data Donation Dataset. LSTM was compared with two naïve forecasting algorithms as well as Ridge linear regression and a random forest using root-mean-square error (RMSE). Parkes error grid quantified clinical accuracy. Regression analysis was used to derive the GVII and GPCI. RESULTS: The LSTM had highest accuracy and best GVII and GPCI. RMSE for CL was 19.8 ± 3.2 and 33.2 ± 5.4 mg/dL for 30- and 60-minute prediction horizons, respectively. RMSE for SAP was 19.6 ± 3.8 and 33.1 ± 7.3 mg/dL for 30- and 60-minute prediction horizons, respectively; 99.6% and 97.6% of predictions were within zones A+B of the Parkes error grid at 30- and 60-minute prediction horizons, respectively. Glucose variability was strongly correlated with RMSE (R≥0.64, P < 0.001); GVII and GPCI demonstrated a means to compare algorithms across datasets with different glucose variability. CONCLUSIONS: The LSTM model was accurate on a large real-world free-living dataset. Glucose variability should be considered when assessing prediction accuracy using indices such as GVII and GPCI.

Comparing fall detection methods in people with multiple sclerosis: A prospective observational cohort study.

Background Falls occur across the population but are more common, and have more negative sequelae, in people with multiple sclerosis (MS). Given the prevalence and impact of falls, accurate measures of fall frequency are needed. This study compares the sensitivity and false discovery rates of three methods of fall detection: the current gold standard, prospective paper fall calendars, real-time self-reporting and automated detection, the latter two from a novel body-worn device. Methods Falls in twenty-five people with MS were recorded for eight weeks with prospective fall calendars, real-time body-worn self-report, and an automated body-worn detector concurrently. Eligible individuals were adults with MS enrolled in a randomized controlled trial of a fall prevention intervention. Entry criteria were at least two falls or near-falls in the previous two months, Expanded Disability Status Scale ≤ 6.0, community dwelling, and no MS relapse in the previous month. The sensitivity (proportion of true falls detected) and false discovery rates (proportion of false reports generated) of the fall detection methods were compared. A true fall was a fall reported by at least two methods. A false report was a fall reported by only one method. The trial is registered on ClinicalTrials.gov (NCT02583386) and is closed. Results In the 1,276 person-days of fall counting with all three methods in use simultaneously there were 1344 unique fall events. Of these, 8.5% (114) were true falls and 91.5% (1230) were false reports. Fall calendars had the lowest sensitivity (0.614) and the lowest false discovery rate (0.067). The automated detector had the highest sensitivity (0.921) and the highest false discovery rate (0.919). All methods generated under one false report per day. There were no fall detection-related adverse events. Conclusion Fall calendars likely underestimate fall frequency by around 40%. The automated detector evaluated here misses very few falls but likely overestimates the number of falls by around one fall per day. Additional research is needed to produce an ideal fall detection and counting method for use in clinical and research applications. Funding United States Department of Veterans Affairs, Rehabilitations Research and Development Service.

Automated Detection of Real-World Falls: Modeled From People With Multiple Sclerosis.

Falls are a major health problem with one in three people over the age of 65 falling each year, oftentimes causing hip fractures, disability, reduced mobility, hospitalization and death. A major limitation in fall detection algorithm development is an absence of real-world falls data. Fall detection algorithms are typically trained on simulated fall data that contain a well-balanced number of examples of falls and activities of daily living. However, real-world falls occur infrequently, making them difficult to capture and causing severe data imbalance. People with multiple sclerosis (MS) fall frequently, and their risk of falling increases with disease progression. Because of their high fall incidence, people with MS provide an ideal model for studying falls. This paper describes the development of a context-aware fall detection system based on inertial sensors and time of flight sensors that is robust to imbalance, which is trained and evaluated on real-world falls in people with MS. The algorithm uses an auto-encoder that detects fall candidates using reconstruction error of accelerometer signals followed by a hyper-ensemble of balanced random forests trained using both acceleration and movement features. On a clinical dataset obtained from 25 people with MS monitored over eight weeks during free-living conditions, 54 falls were observed and our system achieved a sensitivity of 92.14%, and false-positive rate of 0.65 false alarms per day.

Measuring glucose at the site of insulin delivery with a redox-mediated sensor.

Automated insulin delivery systems for people with type 1 diabetes rely on an accurate subcutaneous glucose sensor and an infusion cannula that delivers insulin in response to measured glucose. Integrating the sensor with the infusion cannula would provide substantial benefit by reducing the number of devices inserted into subcutaneous tissue. We describe the sensor chemistry and a calibration algorithm to minimize impact of insulin delivery artifacts in a new glucose sensing cannula. Seven people with type 1 diabetes undergoing automated insulin delivery used two sensing cannulae whereby one delivered a rapidly-acting insulin analog and the other delivered a control phosphate buffered saline (PBS) solution with no insulin. While there was a small artifact in both conditions that increased for larger volumes, there was no difference between the artifacts in the sensing cannula delivering insulin compared with the sensing cannula delivering PBS as determined by integrating the area-under-the-curve of the sensor values following delivery of larger amounts of fluid (P = 0.7). The time for the sensor to recover from the artifact was found to be longer for larger fluid amounts compared with smaller fluid amounts (10.3 ± 8.5 min vs. 41.2 ± 78.3 s, P < 0.05). Using a smart-sampling Kalman filtering smoothing algorithm improved sensor accuracy. When using an all-point calibration on all sensors, the smart-sampling Kalman filter reduced the mean absolute relative difference from 10.9% to 9.5% and resulted in 96.7% of the data points falling within the A and B regions of the Clarke error grid. Despite a small artifact, which is likely due to dilution by fluid delivery, it is possible to continuously measure glucose in a cannula that simultaneously delivers insulin.

An artificial intelligence decision support system for the management of type 1 diabetes.

Type 1 diabetes (T1D) is characterized by pancreatic beta cell dysfunction and insulin depletion. Over 40% of people with T1D manage their glucose through multiple injections of long-acting basal and short-acting bolus insulin, so-called multiple daily injections (MDI)1,2. Errors in dosing can lead to life-threatening hypoglycaemia events (<70 mg dl-1) and hyperglycaemia (>180 mg dl-1), increasing the risk of retinopathy, neuropathy, and nephropathy. Machine learning (artificial intelligence) approaches are being harnessed to incorporate decision support into many medical specialties. Here, we report an algorithm that provides weekly insulin dosage recommendations to adults with T1D using MDI therapy. We employ a unique virtual platform3 to generate over 50,000 glucose observations to train a k-nearest neighbours4 decision support system (KNN-DSS) to identify causes of hyperglycaemia or hypoglycaemia and determine necessary insulin adjustments from a set of 12 potential recommendations. The KNN-DSS algorithm achieves an overall agreement with board-certified endocrinologists of 67.9% when validated on real-world human data, and delivers safe recommendations, per endocrinologist review. A comparison of inter-physician-recommended adjustments to insulin pump therapy indicates full agreement of 41.2% among endocrinologists, which is consistent with previous measures of inter-physician agreement (41-45%)5. In silico3,6 benchmarking using a platform accepted by the United States Food and Drug Administration for evaluation of artificial pancreas technologies indicates substantial improvement in glycaemic outcomes after 12 weeks of KNN-DSS use. Our data indicate that the KNN-DSS allows for early identification of dangerous insulin regimens and may be used to improve glycaemic outcomes and prevent life-threatening complications in people with T1D.

Predicting and Preventing Nocturnal Hypoglycemia in Type 1 Diabetes Using Big Data Analytics and Decision Theoretic Analysis.

Background: Despite new glucose sensing technologies, nocturnal hypoglycemia is still a problem for people with type 1 diabetes (T1D) as symptoms and sensor alarms may not be detected while sleeping. Accurately predicting nocturnal hypoglycemia before sleep may help minimize nighttime hypoglycemia. Methods: A support vector regression (SVR) model was trained to predict, before bedtime, the overnight minimum glucose and overnight nocturnal hypoglycemia for people with T1D. The algorithm was trained on continuous glucose measurements and insulin data collected from 124 people (22,804 valid nights of data) with T1D. The minimum glucose threshold for announcing nocturnal hypoglycemia risk was derived by applying a decision theoretic criterion to maximize expected net benefit. Accuracy was evaluated on a validation set from 10 people with T1D during a 4-week trial under free-living sensor-augmented insulin-pump therapy. The primary outcome measures were sensitivity and specificity of prediction, the correlation between predicted and actual minimum nocturnal glucose, and root-mean-square error. The impact of using the algorithm to prevent nocturnal hypoglycemia is shown in-silico. Results: The algorithm predicted 94.1% of nocturnal hypoglycemia events (<3.9 mmol/L, 95% confidence interval [CI], 71.3-99.9) with an area under the receiver operating characteristic curve of 0.86 (95% CI, 0.75-0.98). Correlation between actual and predicted minimum glucose was high (R = 0.71, P < 0.001). In-silico simulations showed that the algorithm could reduce nocturnal hypoglycemia by 77.0% (P = 0.006) without impacting time in target range (3.9-10 mmol/L). Conclusion: An SVR model trained on a big data set and optimized using decision theoretic criterion can accurately predict at bedtime if overnight nocturnal hypoglycemia will occur and may help reduce nocturnal hypoglycemia.

Design and Evaluation of a Non-ContactBed-Mounted Sensing Device for AutomatedIn-Home Detection of Obstructive Sleep Apnea:A Pilot Study.

We conducted a pilot study to evaluate the accuracy of a custom built non-contactpressure-sensitive device in diagnosing obstructive sleep apnea (OSA) severity as an alternative toin-laboratory polysomnography (PSG) and a Type 3 in-home sleep apnea test (HSAT). Fourteenpatients completed PSG sleep studies for one night with simultaneous recording from ourload-cell-based sensing device in the bed. Subjects subsequently installed pressure sensors in theirbed at home and recorded signals for up to four nights. Machine learning models were optimized toclassify sleep apnea severity using a standardized American Academy of Sleep Medicine (AASM)scoring of the gold standard studies as reference. On a per-night basis, our model reached a correctOSA detection rate of 82.9% (sensitivity = 88.9%, specificity = 76.5%), and OSA severity classificationaccuracy of 74.3% (61.5% and 81.8% correctly classified in-clinic and in-home tests, respectively).There was no difference in Apnea Hypopnea Index (AHI) estimation when subjects wore HSATsensors versus load cells (LCs) only (p-value = 0.62). Our in-home diagnostic system providesan unobtrusive method for detecting OSA with high sensitivity and may potentially be used forlong-term monitoring of breathing during sleep. Further research is needed to address the lowerspecificity resulting from using the highest AHI from repeated samples.

Leveraging a Big Dataset to Develop a Recurrent Neural Network to Predict Adverse Glycemic Events in Type 1 Diabetes.

Patients with type 1 diabetes (T1D) do not produce their own insulin. They must continuously monitor their glucose and make decisions about insulin dosing to avoid the consequences of adverse glucose excursions. Continuous glucose monitoring (CGM) systems and insulin pumps are state-of-the-art systems that can help people with T1D manage their glucose. Accurate glucose prediction algorithms are becoming critical components of CGM systems that can help people with T1D proactively avoid the occurrence of impending hyperglycemia and hypoglycemia events. We present Glucop30, a robust data-driven glucose prediction model that is trained on a big dataset (27,466 days) to forecast glucose concentration along a short-term prediction horizon of 30 minutes. Our proposed prediction method is composed of (i) a recurrent neural network with long-short-term-memory (LSTM) units that predicts the general trend of future glucose levels, followed by (ii) a patient-specific smoothing error correction step that accounts for inter- and intra-patient glucose variability. We retrospectively test our proposed method on a clinical dataset obtained from 10 T1D insulin pump users who were continuously monitored during a 4-week trial under free-living conditions (255 days), and assess the impact of the size of the training set on the accuracy of the proposed model. In addition, we report on the accuracy of our method when both CGM and insulin data are used for prediction; however we discovered that adding insulin as an additional input feature improves prediction accuracy by only 1%. Glucop30 achieves leading performance as measured by the RMSE of 7.55 (SD = 2.20 mg/dL) and MAE of 4.89 (SD = 1.43 mg/dL) for an effective prediction horizon of 27.50 (SD = 2.64) minutes. Moreover, Glucop30 accurately anticipates the occurrence of 97.79 (SD = 5.35)% of hyperglycemia events (glucose > 180 mg/dL), and 90.87 (SD = 6.79)% of hypoglycemia events (glucose < 70 mg/dL) with remarkably few false alerts (1 and 2 false alarms per week for hypoglycemia and hyperglycemia events, respectively).

In-Home Sleep Apnea Severity Classification using Contact-free Load Cells and an AdaBoosted Decision Tree Algorithm.

We present a method for automated diagnosis and classification of severity of sleep apnea using an array of non-contact pressure-sensitive sensors placed underneath a mattress as an alternative to conventional obtrusive sensors. Our algorithm comprises two stages: i) A decision tree classifier that identifies patients with sleep apnea, and ii) a subsequent linear regression model that estimates the Apnea-Hypopnea Index (AHI), which is used to determine the severity of sleep disordered breathing. We tested our algorithm on a cohort of 14 patients who underwent overnight home sleep apnea test. The machine learning algorithm was trained and performance was evaluated using leave-one-patient-out cross-validation. The accuracy of the proposed approach in detecting sleep apnea is 86.96%, with sensitivity and specificity of 81.82% and 91.67%, respectively. Moreover, classification of severity of the sleep disorder was correctly assigned in 11 out of 14 cases, and the mean absolute error in the AHI estimation was calculated to be 3.83 events/hr.

Design and evaluation of a portable smart-phone based peripheral neuropathy test platform.

In this paper, we describe a novel portable test platform that can be used to test peripheral neuropathy either within a clinic or at home. The system, called the PeriVib, is comprised of (1) a small, custom vibration motor designed to apply a vibration stimulus to the toe with constant pressure to test sensation threshold, and (2) a custom smart-phone app that enables a patient to run a series of functional gait and balance tests. Vibration is applied by PeriVib in two separate modes. The first mode, ramp-up, starts at zero amplitude and increases to a maximum level while the patient indicates when they start feeling the pressure by lifting their finger off the touch-screen on the phone. The second mode, ramp-down, starts at a maximal intensity and decreases in intensity; the patient indicates when they stop feeling the vibration. The smart-phone app determines the patient's threshold by recording the vibration amplitude when they indicate the onset or loss of vibratory sensation, depending on the mode. In both modes, the measurement is repeated five times. In addition to controlling the vibration motor during the vibration test, the smart phone app also enables collection of gait and sway metrics through the use of the accelerometer and gyroscope sensors on the smartphone. The entire set of tests requires approximately 5 minutes to complete and can be done by a patient with minimal instructions from a clinician. In a cohort of 28 subjects with a history of chemotherapy-induced peripheral neuropathy, we compared the PeriVib performance with two established threshold sensing systems: (1) a Biothesiometer device and (2) a tuning fork. We found that the sensation threshold estimated by PeriVib correlated well with the Biothesiometer ($\mathrm{R}^{2}$ of 0.68) but less well with the tuning fork ($\mathrm{R}^{2}$ of 0.15). Functional gait and balance metrics did not correlate with peripheral neuropathy severity.