Evaluating inter-individual variability captured by the Eleveld pharmacokinetics model.

Inter-individual variability in Pharmacokinetic (PK) and Pharmacodynamic (PD) models significantly affects the accuracy of Target Controlled Infusion and closed-loop control of anesthesia. We hypothesize that the novel Eleveld PK model captures more inter-individual variability relevant to both open-loop and closed-loop control design, resulting in reduced variability in PD models identified using the Eleveld PK model's plasma prediction compared to the Schuttler or Schnider PK model. We used a dataset of propofol infusion rates and Depth of Hypnosis measurements across three demographic groups: elderly, obese, and adult. PD models are identified based on plasma concentration prediction using three PK models (Schuttler, Schnider, and Eleveld). Validation methods are presented to confirm acceptable predictive performance and comparable PK-PD model variability within each demographic group. To test our hypothesis, we compared coefficient variations in step responses for open-loop control and multiplicative uncertainty of PD model sets for closed-loop control. Validated PKPD models using the Schuttler and Schnider PK model showed no significant differences in predictive response and multiplicative uncertainty compared to the Eleveld PK model. The coefficient variations in step responses of PD model sets and the frequency ranges, corresponding to uncertainty below one, were comparable for all three PK models. The comparison of the accumulated coefficient of variation in the step-response and the uncertainty of the PD model sets indicated that the Eleveld PK model does not offer any advantage for the design of open-loop or closed-loop control of anesthesia.

Detection of spreading depression features from the scalp of epileptic patients.

Spreading depression (SD), a pathological cortical negative DC potential, is caused by an elevation of potassium ions in the extracellular space. This leads to a transient relocation of ions within neurons and a slow spread through brain tissue. Our previous research established a correlation between scalp SD and seizures in patients with intractable epilepsy using our novel electroencephalography (EEG). In this study, we enhanced our EEG system by incorporating a Near-infrared spectroscopy (NIRS) module for multi-modal EEG-NIRS measurements. The aim is to provide an investigation into the defining characteristics and methods for detecting SD.Clinical Relevance-: The detection of SD serves as a novel biomarker for epilepsy, capable of forewarning seizures within a time range from 10 secs to 30 min. This detection plays a crucial role in predicting and preventing seizures and providing diagnostic information for drug-resistant epilepsy patients.

Neonatal Face and Facial Landmark Detection from Video Recordings.

This paper explores automated face and facial landmark detection of neonates, which is an important first step in many video-based neonatal health applications, such as vital sign estimation, pain assessment, sleep-wake classification, and jaundice detection. Utilising three publicly available datasets of neonates in the clinical environment, 366 images (258 subjects) and 89 (66 subjects) were annotated for training and testing, respectively. Transfer learning was applied to two YOLO-based models, with input training images augmented with random horizontal flipping, photo-metric colour distortion, translation and scaling during each training epoch. Additionally, the re-orientation of input images and fusion of trained deep learning models was explored. Our proposed model based on YOLOv7Face outperformed existing methods with a mean average precision of 84.8% for face detection, and a normalised mean error of 0.072 for facial landmark detection. Overall, this will assist in the development of fully automated neonatal health assessment algorithms.Clinical relevance- Accurate face and facial landmark detection provides an automated and non-contact option to assist in video-based neonatal health applications.

Blink-related EEG oscillations are neurophysiological indicators of subconcussive head impacts in female soccer players: a preliminary study.

Introduction: Repetitive subconcussive head impacts can lead to subtle neural changes and functional consequences on brain health. However, the objective assessment of these changes remains limited. Resting state blink-related oscillations (BROs), recently discovered neurological responses following spontaneous blinking, are explored in this study to evaluate changes in BRO responses in subconcussive head impacts. Methods: We collected 5-min resting-state electroencephalography (EEG) data from two cohorts of collegiate athletes who were engaged in contact sports (SC) or non-contact sports (HC). Video recordings of all on-field activities were conducted to determine the number of head impacts during games and practices in the SC group. Results: In both groups, we were able to detect a BRO response. Following one season of games and practice, we found a strong association between the number of head impacts sustained by the SC group and increases in delta and beta spectral power post-blink. There was also a significant difference between the two groups in the morphology of BRO responses, including decreased peak-to-peak amplitude of response over left parietal channels and differences in spectral power in delta and alpha frequency range post-blink. Discussion: Our preliminary results suggest that the BRO response may be a useful biomarker for detecting subtle neural changes resulting from repetitive head impacts. The clinical utility of this biomarker will need to be validated through further research with larger sample sizes, involving both male and female participants, using a longitudinal design.

Concurrent recordings of slow DC-potentials and epileptiform discharges: Novel EEG amplifier and signal processing techniques.

Ionic currents within the brain generate voltage oscillations. These bioelectrical activities include ultra-low frequency electroencephalograms (DC-EEG, frequency less than 0.1 Hz) and conventional clinical electroencephalograms (AC-EEG, 0.5-70 Hz). Although AC-EEG is commonly used for diagnosing epilepsy, recent studies indicate that DC-EEG is an essential frequency component of EEG and can provide valuable information for analyzing epileptiform discharges. During conventional EEG recordings, DC-EEG is censored by applying high-pass filtering to i) obliterate slow-wave artifacts, ii) eliminate the bioelectrodes' half-cell potential asymmetrical changes in ultralow-low frequency, and iii) prevent instrument saturation. Spreading depression (SD), which is the most prolonged fluctuation in DC-EEG, may be associated with epileptiform discharges. However, recording of SD signals from the scalp's surface can be challenging due to the filtering effect and non-neuronal slow shift potentials. In this study, we describe a novel technique to extend the frequency bandwidth of surface EEG to record SD signals. The method includes novel instrumentation, appropriate bioelectrodes, and efficient signal-processing techniques. To evaluate the accuracy of our approach, we performed a simultaneous surface recording of DC- and AC-EEG from epileptic patients during long-term video EEG monitoring, which provide a promising tool for diagnosis of epilepsy. DATA AVAILABILITY STATEMENT: The data presented in this study are available on request.

Choosing blood pressure thresholds to inform pregnancy care in the community: An analysis of cluster trials.

OBJECTIVE: To inform digital health design by evaluating diagnostic test properties of antenatal blood pressure (BP) outputs and levels to identify women at risk of adverse outcomes. DESIGN: Planned secondary analysis of cluster randomised trials. SETTING: India, Pakistan, Mozambique. POPULATION: Women with in-community BP measurements and known pregnancy outcomes. METHODS: Blood pressure was defined by its outputs (systolic and/or diastolic, systolic only, diastolic only or mean arterial pressure [calculated]) and level: normotension-1 (<135/85 mmHg), normotension-2 (135-139/85-89 mmHg), non-severe hypertension (140-149/90-99 mmHg; 150-154/100-104 mmHg; 155-159/105-109 mmHg) and severe hypertension (≥160/110 mmHg). Dose-response (adjusted risk ratio [aRR]) and diagnostic test properties (negative [-LR] and positive [+LR] likelihood ratios) were estimated. MAIN OUTCOME MEASURES: Maternal/perinatal composites of mortality/morbidity. RESULTS: Among 21 069 pregnancies, different BP outputs had similar aRR, -LR, and +LR for adverse outcomes. No BP level (even normotension-1) was associated with low risk (all -LR ≥0.20). Across outcomes, risks rose progressively with higher BP levels above normotension-1. For each of maternal central nervous system events and stillbirth, BP ≥155/105 mmHg showed at least good diagnostic test performance (+LR ≥5.0) and BP ≥135/85 mmHg at least fair performance, similar to BP ≥140/90 mmHg (+LR 2.0-4.99). CONCLUSIONS: In the community, normal BP values do not provide reassurance about subsequent adverse outcomes. Given the similar performance of BP cut-offs of 135/85 and 140/90 mmHg for hypertension, and 155/105 and 160/110 mmHg for severe hypertension, digital decision support for women in the community should consider using these lower thresholds.

Effective pandemic policy design through feedback does not need accurate predictions.

The COVID-19 pandemic has had an enormous toll on human health and well-being and led to major social and economic disruptions. Public health interventions in response to burgeoning case numbers and hospitalizations have repeatedly bent down the epidemic curve, effectively creating a feedback control system. Worst case scenarios have been avoided in many places through this responsive feedback. We aim to formalize and illustrate how to incorporate principles of feedback control into pandemic projections and decision-making, and ultimately shift the focus from prediction to the design of interventions. Starting with an epidemiological model for COVID-19, we illustrate how feedback control can be incorporated into pandemic management using a simple design that couples recent changes in case numbers or hospital occupancy with explicit policy restrictions. We demonstrate robust ability to control a pandemic using a design that responds to hospital cases, despite simulating large uncertainty in reproduction number R0 (range: 1.04-5.18) and average time to hospital admission (range: 4-28 days). We show that shorter delays, responding to case counts versus hospital measured infections, reduce both the cumulative case count and the average level of interventions. Finally, we show that feedback is robust to changing compliance to public health directives and to systemic changes associated with variants of concern and with the introduction of a vaccination program. The negative impact of a pandemic on human health and societal disruption can be reduced by coupling models of disease propagation with models of the decision-making process. In contrast to highly varying open-loop projections, incorporating feedback explicitly in the decision-making process is more reflective of the real-world challenge facing public health decision makers. Using feedback principles, effective control strategies can be designed even if the pandemic characteristics are highly uncertain, encouraging earlier and smaller actions that reduce both case counts and the extent of interventions.

Prediction of Neonatal Respiratory Distress in Term Babies at Birth from Digital Stethoscope Recorded Chest Sounds.

Neonatal respiratory distress is a common condition that if left untreated, can lead to short- and long-term complications. This paper investigates the usage of digital stethoscope recorded chest sounds taken within 1 min post-delivery, to enable early detection and prediction of neonatal respiratory distress. Fifty-one term newborns were included in this study, 9 of whom developed respiratory distress. For each newborn, 1 min anterior and posterior recordings were taken. These recordings were pre-processed to remove noisy segments and obtain high-quality heart and lung sounds. The random undersampling boosting (RUSBoost) classifier was then trained on a variety of features, such as power and vital sign features extracted from the heart and lung sounds. The RUSBoost algorithm produced specificity, sensitivity, and accuracy results of 85.0%, 66.7% and 81.8%, respectively. Clinical relevance--- This paper investigates the feasibility of digital stethoscope recorded chest sounds for early detection of respiratory distress in term newborn babies, to enable timely treatment and management.

A New Non-Negative Matrix Co-Factorisation Approach for Noisy Neonatal Chest Sound Separation.

Obtaining high quality heart and lung sounds enables clinicians to accurately assess a newborns cardio-respiratory health and provide timely care. However, noisy chest sound recordings are common, hindering timely and accurate assessment. A new Non-negative Matrix Co-Factorisation based approach is proposed to separate noisy chest sound recordings into heart, lung and noise components to address this problem. This method is achieved through training with 20 high quality heart and lung sounds, in parallel with separating the sounds of the noisy recording. The method was tested on 68 10-second noisy recordings containing both heart and lung sounds and compared to the current state of the art Non-negative Matrix Factorisation methods. Results show significant improvements in heart and lung sound quality scores respectively, and improved accuracy of 3.6bpm and 1.2bpm in heart and breathing rate estimation respectively, when compared to existing methods.

Model-plant mismatch detection for cross-directional processes.

This paper presents a two-component framework to detect model-plant mismatch (MPM) in cross-directional (CD) processes on paper machines under model-predictive control. First, routine operating data is used for system identification in closed loop; second, a one-class support vector machine (SVM) is trained to predict MPM. The iterative identification method alternates between identifying the finite impulse response coefficients of the spatial and temporal models. It converges, and the parameter estimates are asymptotically consistent. Coefficient estimates drawn from normal operation are used to train a one-class SVM, which then detects model-plant mismatch in subsequent routine operation. This approach applies to routine operating data without requiring external excitations. It can also distinguish mismatches in the process model from changes in the noise model. Examples of CD processes on paper machines are provided to verify the effectiveness of both components.

The Effect of Low-Dose Intraoperative Ketamine on Closed-Loop-Controlled General Anesthesia: A Randomized Controlled Equivalence Trial.

BACKGROUND: Closed-loop control of propofol-remifentanil anesthesia using the processed electroencephalography depth-of-hypnosis index provided by the NeuroSENSE monitor (WAVCNS) has been previously described. The purpose of this placebo-controlled study was to evaluate the performance (percentage time within ±10 units of the setpoint during the maintenance of anesthesia) of a closed-loop propofol-remifentanil controller during induction and maintenance of anesthesia in the presence of a low dose of ketamine. METHODS: Following ethical approval and informed consent, American Society of Anesthesiologist (ASA) physical status I-II patients aged 19-54 years, scheduled for elective orthopedic surgery requiring general anesthesia for >60 minutes duration, were enrolled in a double-blind randomized, placebo-controlled, 2-group equivalence trial. Immediately before induction of anesthesia, participants in the ketamine group received a 0.25 mg·kg-1 bolus of intravenous ketamine over 60 seconds followed by a continuous 5 µg·kg-1·min-1 infusion for up to 45 minutes. Participants in the control group received an equivalent volume of normal saline. After the initial study drug bolus, closed-loop induction of anesthesia was initiated; propofol and remifentanil remained under closed-loop control until the anesthetic was tapered and turned off at the anesthesiologist's discretion. An equivalence range of ±8.99% was assumed for comparing controller performance. RESULTS: Sixty patients participated: 41 males, 54 ASA physical status I, with a median (interquartile range [IQR]) age of 29 [23, 38] years and weight of 82 [71, 93] kg. Complete data were available from 29 cases in the ketamine group and 27 in the control group. Percentage time within ±10 units of the WAVCNS setpoint was median [IQR] 86.6% [79.7, 90.2] in the ketamine group and 86.4% [76.5, 89.8] in the control group (median difference, 1.0%; 95% confidence interval [CI] -3.6 to 5.0). Mean propofol dose during maintenance of anesthesia for the ketamine group was higher than for the control group (median difference, 24.9 µg·kg-1·min-1; 95% CI, 6.5-43.1; P = .005). CONCLUSIONS: Because the 95% CI of the difference in controller performance lies entirely within the a priori equivalence range, we infer that this analgesic dose of ketamine did not alter controller performance. Further study is required to confirm the finding that mean propofol dosing was higher in the ketamine group, and to investigate the implication that this dose of ketamine may have affected the WAVCNS.

Effect of ketamine on the NeuroSENSE WAVCNS during propofol anesthesia; a randomized feasibility trial.

Dose-dependent effects of ketamine on processed electroencephalographic depth-of-hypnosis indices have been reported. Limited data are available for the NeuroSENSE WAVCNS index. Our aim was to establish the feasibility of closed-loop propofol-remifentanil anesthesia guided by the WAVCNS index in the presence of an analgesic dose of ketamine. Thirty ASA I-II adults, 18-54 years, requiring general anesthesia for anterior cruciate ligament surgery were randomized to receive: full-dose [ketamine, 0.5 mg kg-1 initial bolus, 10 mcg kg-1 min-1 infusion] (recommended dose for postoperative pain management); half-dose [ketamine, 0.25 mg kg-1 bolus, 5 mcg kg-1 min-1 infusion]; or control [no ketamine]. After the ketamine bolus, patients received 1.0 mcg kg-1 remifentanil over 30 s, then 1.5 mg kg-1 propofol over 30 s, followed by manually-adjusted propofol-remifentanil anesthesia. The WAVCNS was > 60 for 7/9 patients in the full-dose group at 7 min after starting the propofol infusion. This was inconsistent with clinical observations of depth-of-hypnosis and significantly higher than control (median difference [MD] 17.0, 95% confidence interval [CI] 11.4-26.8). WAVCNS was median [interquartile range] 49.3 [42.2-62.6] in the half-dose group, and not different to control (MD 5.1, 95% CI - 4.9 to 17.9). During maintenance of anesthesia, the WAVCNS was higher in the full-dose group compared to control (MD 14.7, 95% CI 10.2-19.2) and in the half-dose group compared to control (MD 11.4, 95% CI 4.7-20.4). The full-dose of ketamine recommended for postoperative pain management had a significant effect on the WAVCNS. This effect should be considered when using the WAVCNS to guide propofol-remifentanil dosing.Trial Registration ClinicalTrails.gov No. NCT02908945.

Neonatal Heart and Lung Sound Quality Assessment for Robust Heart and Breathing Rate Estimation for Telehealth Applications.

With advances in digital stethoscopes, internet of things, signal processing and machine learning, chest sounds can be easily collected and transmitted to the cloud for remote monitoring and diagnosis. However, low quality of recordings complicates remote monitoring and diagnosis, particularly for neonatal care. This paper proposes a new method to objectively and automatically assess the signal quality to improve the accuracy and reliability of heart rate (HR) and breathing rate (BR) estimation from noisy neonatal chest sounds. A total of 88 10-second long chest sounds were taken from 76 preterm and full-term babies. Six annotators independently assessed the signal quality, number of detectable beats, and breathing periods from these recordings. For quality classification, 187 and 182 features were extracted from heart and lung sounds, respectively. After feature selection, class balancing, and hyperparameter optimization, a dynamic binary classification model was trained. Then HR and BR were automatically estimated from the chest sound and several approaches were compared.The results of subject-wise leave-one-out cross-validation, showed that the model distinguished high and low quality recordings in the test set with 96% specificity, 81% sensitivity and 93% accuracy for heart sounds, and 86% specificity, 69% sensitivity and 82% accuracy for lung sounds. The HR and BR estimated from high quality sounds resulted in significantly less median absolute error (4 bpm and 12 bpm difference, respectively) compared to those from low quality sounds. The methods presented in this work, facilitates automated neonatal chest sound auscultation for future telehealth applications.

Robust PID control of propofol anaesthesia: Uncertainty limits performance, not PID structure.

BACKGROUND AND OBJECTIVE: New proposals to improve the regulation of hypnosis in anaesthesia based on the development of advanced control structures emerge continuously. However, a fair study to analyse the real benefits of these structures compared to simpler clinically validated PID-based solutions has not been presented so far. The main objective of this work is to analyse the performance limitations associated with using a filtered PID controller, as compared to a high-order controller, represented through a Youla parameter. METHODS: The comparison consists of a two-steps methodology. First, two robust optimal filtered PID controllers, considering the effect of the inter-patient variability, are synthesised. A set of 47 validated paediatric pharmacological models, identified from clinical data, is used to this end. This model set provides representative inter-patient variability Second, individualised filtered PID and Youla controllers are synthesised for each model in the set. For fairness of comparison, the same performance objective is optimised for all designs, and the same robustness constraints are considered. Controller synthesis is performed utilising convex optimisation and gradient-based methods relying on algebraic differentiation. The worst-case performance over the patient model set is used for the comparison. RESULTS: Two robust filtered PID controllers for the entire model set, as well as individual-specific PID and Youla controllers, were optimised. All considered designs resulted in similar frequency response characteristics. The performance improvement associated with the Youla controllers was not significant compared to the individually tuned filtered PID controllers. The difference in performance between controllers synthesized for the model set and for individual models was significantly larger than the performance difference between the individual-specific PID and Youla controllers. The different controllers were evaluated in simulation. Although all of them showed clinically acceptable results, the robust solutions provided slower responses. CONCLUSION: Taking the same clinical and technical considerations into account for the optimisation of the different controllers, the design of individual-specific solutions resulted in only marginal differences in performance when comparing an optimal Youla parameter and its optimal filtered PID counterpart. The inter-patient variability is much more detrimental to performance than the limitations imposed by the simple structure of the filtered PID controller.

Association of cortical spreading depression and seizures in patients with medically intractable epilepsy.

OBJECTIVE: Monitoring of the ultra-low frequency potentials, particularly cortical spreading depression (CSD), is excluded in epilepsy monitoring due to technical barriers imposed by the scalp ultra-low frequency electroencephalogram (EEG). As a result, clinical studies of CSD have been limited to invasive EEG. Therefore, the occurrence of CSD and its interaction with epileptiform field potentials (EFP) require investigation in epilepsy monitoring. METHODS: Using a novel AC/DC-EEG approach, the occurrence of DC potentials in patients with intractable epilepsy presenting different symptoms of aura was investigated during long-term video-EEG monitoring. RESULTS: Various forms of slow potentials, including simultaneous negative direct current (DC) potentials and prolonged EFP, propagated negative DC potentials, and non-propagated single negative DC potentials were recorded from the scalp of the epileptic patients. The propagated and single negative DC potentials preceded the prolonged EFP with a time lag and seizure appeared at the final shoulder of some instances of the propagated negative DC potentials. The slow potential deflections had a high amplitude and prolonged duration and propagated slowly through the brain. The high-frequency EEG was suppressed in the vicinity of the negative DC potential propagations. CONCLUSIONS: The study is the first to report the recording of the propagated and single negative DC potentials with EFP at the scalp of patients with intractable epilepsy. The negative DC potentials preceded the prolonged EFP and may trigger seizures. The propagated and single negative DC potentials may be considered as CSD. SIGNIFICANCE: Recordings of CSD may serve as diagnostic and prognostic monitoring tools in epilepsy.

Individualized closed-loop anesthesia through patient model partitioning.

Closed-loop controlled drug dosing has the potential of revolutionizing clinical anesthesia. However, inter-patient variability in drug sensitivity poses a central challenge to the synthesis of safe controllers. Identifying a full individual pharmacokinetic-pharmacodynamic (PKPD) model for this synthesis is clinically infeasible due to limited excitation of PKPD dynamics and presence of unmodeled disturbances. This work presents a novel method to mitigate inter-patient variability. It is based on: 1) partitioning an a priori known model set into subsets; 2) synthesizing an optimal robust controller for each subset; 3) classifying patients into one of the subsets online based on demographic or induction phase data; 4) applying the associated closed-loop controller. The method is investigated in a simulation study, utilizing a set of 47 clinically obtained patient models. Results are presented and discussed.Clinical relevance-The proposed method is easy to implement in clinical practice, and has potential to reduce the impact from surgical stimulation disturbances, and to result in safer closed-loop anesthesia with less risk of under- and over dosing.

Evaluating NeuroSENSE for assessing depth of hypnosis during desflurane anesthesia: an adaptive, randomized-controlled trial.

PURPOSE: Processed electroencephalography (EEG) monitors support depth-of-hypnosis assessment during anesthesia. This randomized-controlled trial investigated the performance of the NeuroSENSE electroencephalography (EEG) monitor to determine whether its wavelet anesthetic value for central nervous system (WAVCNS) index distinguishes consciousness from unconsciousness during induction of anesthesia (as assessed by the anesthesiologist) and emergence from anesthesia (indicated by patient responsiveness), and whether it correlates with changes in desflurane minimum alveolar concentration (MAC) during maintenance of anesthesia. METHODS: EEG was collected using a fronto-temporal bilateral montage. The WAVCNS was continuously recorded by the NeuroSENSE monitor, to which the anesthesiologist was blinded. Anesthesia was induced with propofol/remifentanil and maintained with desflurane, with randomized changes of -0.4, 0, or +0.4 MAC every 7.5 min within the 0.8-1.6 MAC range, if clinically acceptable to the anesthesiologist. During emergence from anesthesia, desflurane was stepped down by 0.2 MAC every five minutes. RESULTS: Data from 75 patients with a median [interquartile range] age of 41[35-52] yr were obtained. The WAVCNS distinguished consciousness from unconsciousness as assessed by the anesthesiologist, with area under the receiver operating characteristic curve of 99.5% (95% confidence interval [CI], 98.5 to 100.0) at loss of consciousness and 99.4% (95% CI, 98.5 to 100.0) at return of consciousness. Bilateral WAVCNS changes correlated with desflurane concentrations, with -8.0 and -8.6 WAVCNS units, respectively, per 1 MAC change in the 0.8-1.6 MAC range during maintenance of anesthesia and -10.0 and -10.5 WAVCNS units, respectively, in the 0.4-1.6 MAC range including emergence from anesthesia. CONCLUSION: The NeuroSENSE monitor can reliably discriminate between consciousness and unconsciousness, as assessed by the anesthesiologist, during induction of anesthesia and with a lower level of reliability during emergence from anesthesia. The WAVCNS correlates with desflurane concentration but plateaus at higher concentrations, similar to other EEG monitors, which suggests limited utility to titrate higher concentrations of anesthetic vapour. TRIAL REGISTRATION: clinicaltrials.gov, NCT02088671; registered 17 March, 2014.

Pediatric pulse oximetry-based OSA screening at different thresholds of the apnea-hypopnea index with an expression of uncertainty for inconclusive classifications.

BACKGROUND: Assessments of pediatric obstructive sleep apnea (OSA) are underutilized across Canada due to a lack of resources. Polysomnography (PSG) measures OSA severity through the average number of apnea/hypopnea events per hour (AHI), but is resource intensive and requires a specialized sleep laboratory, which results in long waitlists and delays in OSA detection. Prompt diagnosis and treatment of OSA are crucial for children, as untreated OSA is linked to behavioral deficits, growth failure, and negative cardiovascular consequences. We aim to assess the performance of a portable pediatric OSA screening tool at different AHI cut-offs using overnight smartphone-based pulse oximetry. MATERIAL AND METHODS: Following ethics approval and informed consent, children referred to British Columbia Children's Hospital for overnight PSG were recruited for two studies including 160 and 75 children, respectively. An additional smartphone-based pulse oximeter sensor was used in both studies to record overnight pulse oximetry [SpO2 and photoplethysmogram (PPG)] alongside the PSG. Features characterizing SpO2 dynamics and heart rate variability from pulse peak intervals of the PPG signal were derived from pulse oximetry recordings. Three multivariate logistic regression screening models, targeted at three different levels of OSA severity (AHI ≥ 1, 5, and 10), were developed using stepwise-selection of features using the Bayesian information criterion (BIC). The "Gray Zone" approach was also implemented for different tolerance values to allow for more precise detection of children with inconclusive classification results. RESULTS: The optimal diagnostic tolerance values defining the "Gray Zone" borders (15, 10, and 5, respectively) were selected to develop the final models to screen for children at AHI cut-offs of 1, 5, and 10. The final models evaluated through cross-validation showed good accuracy (75%, 82% and 89%), sensitivity (80%, 85% and 82%) and specificity (65%, 79% and 91%) values for detecting children with AHI ≥ 1, AHI ≥ 5 and AHI ≥ 10. The percentage of children classified as inconclusive was 28%, 38% and 16% for models detecting AHI ≥ 1, AHI ≥ 5, and AHI ≥ 10, respectively. CONCLUSIONS: The proposed pulse oximetry-based OSA screening tool at different AHI cut-offs may assist clinicians in identifying children at different OSA severity levels. Using this tool at home prior to PSG can help with optimizing the limited resources for PSG screening. Further validation with larger and more heterogeneous datasets is required before introducing in clinical practice.