Dataset Information

Supervised Machine Learning Applied to Automate Flash and Prolonged Capillary Refill Detection by Pulse Oximetry.

ABSTRACT: Objective:Develop an automated approach to detect flash (<1.0 s) or prolonged (>2.0 s) capillary refill time (CRT) that correlates with clinician judgment by applying several supervised machine learning (ML) techniques to pulse oximeter plethysmography data. Materials and Methods:Data was collected in the Pediatric Intensive Care Unit (ICU), Cardiac ICU, Progressive Care Unit, and Operating Suites in a large academic children's hospital. Ninety-nine children and 30 adults were enrolled in testing and validation cohorts, respectively. Patients had 5 paired CRT measurements by a modified pulse oximeter device and a clinician, generating 485 waveform pairs for model training. Supervised ML models using gradient boosting (XGBoost), logistic regression (LR), and support vector machines (SVMs) were developed to detect flash (<1 s) or prolonged CRT (?2 s) using clinician CRT assessment as the reference standard. Models were compared using Area Under the Receiver Operating Curve (AUC) and precision-recall curve (positive predictive value vs. sensitivity) analysis. The best performing model was externally validated with 90 measurement pairs from adult patients. Feature importance analysis was performed to identify key waveform characteristics. Results:For flash CRT, XGBoost had a greater mean AUC (0.79, 95% CI 0.75-0.83) than logistic regression (0.77, 0.71-0.82) and SVM (0.72, 0.67-0.76) models. For prolonged CRT, XGBoost had a greater mean AUC (0.77, 0.72-0.82) than logistic regression (0.73, 0.68-0.78) and SVM (0.75, 0.70-0.79) models. Pairwise testing showed statistically significant improved performance comparing XGBoost and SVM; all other pairwise model comparisons did not reach statistical significance. XGBoost showed good external validation with AUC of 0.88. Feature importance analysis of XGBoost identified distinct key waveform characteristics for flash and prolonged CRT, respectively. Conclusion:Novel application of supervised ML to pulse oximeter waveforms yielded multiple effective models to identify flash and prolonged CRT, using clinician judgment as the reference standard. Tweet:Supervised machine learning applied to pulse oximeter waveform features predicts flash or prolonged capillary refill.

SUBMITTER: Hunter RB

PROVIDER: S-EPMC7574820 | biostudies-literature | 2020

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Supervised Machine Learning Applied to Automate Flash and Prolonged Capillary Refill Detection by Pulse Oximetry.

Hunter Ryan Brandon RB Jiang Shen S Nishisaki Akira A Nickel Amanda J AJ Napolitano Natalie N Shinozaki Koichiro K Li Timmy T Saeki Kota K Becker Lance B LB Nadkarni Vinay M VM Masino Aaron J AJ

Frontiers in physiology 20201006

<h4>Objective</h4>Develop an automated approach to detect flash (<1.0 s) or prolonged (>2.0 s) capillary refill time (CRT) that correlates with clinician judgment by applying several supervised machine learning (ML) techniques to pulse oximeter plethysmography data.<h4>Materials and methods</h4>Data was collected in the Pediatric Intensive Care Unit (ICU), Cardiac ICU, Progressive Care Unit, and Operating Suites in a large academic children's hospital. Ninety-nine children and 30 adults were enr ...[more]

PMID: 33117190

Similar Datasets

Project description:BackgroundsIn children with sepsis, circulatory shock and multi-organ failure remain major contributors to mortality. Prolonged capillary refill time (PCRT) is a clinical tool associated with disease severity and tissue hypoperfusion. Microcirculation assessment with videomicroscopy represents a promising candidate for assessing and improving hemodynamic management strategies in children with sepsis. Particularly when there is loss of coherence between the macro and microcirculation (hemodynamic incoherence). We sought to evaluate the association between PCRT and microcirculation changes in sepsis.MethodsThis was a prospective cohort study in children hospitalized with sepsis. Microcirculation was measured using sublingual video microscopy (capillary density and flow and perfused boundary region [PBR]-a parameter inversely proportional to vascular endothelial glycocalyx thickness), phalangeal tissue perfusion, and endothelial activation and glycocalyx injury biomarkers. The primary outcome was the association between PCRT and microcirculation changes.ResultsA total of 132 children with sepsis were included, with a median age of two years (IQR 0.6-12.2). PCRT was associated with increased glycocalyx degradation (PBR 2.21 vs. 2.08 microns; aOR 2.65, 95% CI 1.09-6.34; p = 0.02) and fewer 4-6 micron capillaries recruited (p = 0.03), with no changes in the percentage of capillary blood volume (p = 0.13). Patients with hemodynamic incoherence had more PBR abnormalities (78.4% vs. 60.8%; aOR 2.58, 95% CI 1.06-6.29; p = 0.03) and the persistence of these abnormalities after six hours was associated with higher mortality (16.5% vs. 6.1%; p < 0.01). Children with an elevated arterio-venous CO2 difference (DCO2) had an abnormal PBR (aOR 1.13, 95% CI 1.01-1.26; p = 0.03) and a lower density of small capillaries (p < 0.05). Prolonged capillary refill time predicted an abnormal PBR (AUROC 0.81, 95% CI 0.64-0.98; p = 0.03) and relative percentage of blood in the capillaries (AUROC 0.82, 95% CI 0.58-1.00; p = 0.03) on admission. A normal CRT at 24 h predicted a shorter hospital stay (aOR 0.96, 95% CI 0.94-0.99; p < 0.05).ConclusionsWe found an association between PCRT and microcirculation changes in children with sepsis. These patients had fewer small capillaries recruited and more endothelial glycocalyx degradation. This leads to nonperfused capillaries, affecting oxygen delivery to the tissues. These disorders were associated with hemodynamic incoherence and worse clinical outcomes when the CRT continued to be abnormal 24 h after admission.

Project description:BackgroundA peripheral perfusion-targeted resuscitation during early septic shock has shown encouraging results. Capillary refill time, which has a prognostic value, was used. Adding accuracy and predictability on capillary refill time (CRT) measurement, if feasible, would benefit to peripheral perfusion-targeted resuscitation. We assessed whether a reduction of capillary refill time during passive leg raising (ΔCRT-PLR) predicted volume-induced peripheral perfusion improvement defined as a significant decrease of capillary refill time following volume expansion.MethodsThirty-four patients with acute circulatory failure were selected. Haemodynamic variables, metabolic variables (PCO2gap), and four capillary refill time measurements were recorded before and during a passive leg raising test and after a 500-mL volume expansion over 20 min. Receiver operating characteristic curves were built, and areas under the curves were calculated (ROCAUC). Confidence intervals (CI) were performed using a bootstrap analysis. We recorded mortality at day 90.ResultsThe least significant change in the capillary refill time was 25% [95% CI, 18-30]. We defined CRT responders as patients showing a reduction of at least 25% of capillary refill time after volume expansion. A decrease of 27% in ΔCRT-PLR predicted peripheral perfusion improvement with a sensitivity of 87% [95% CI, 73-100] and a specificity of 100% [95% CI, 74-100]. The ROCAUC of ΔCRT-PLR was 0.94 [95% CI, 0.87-1.0]. The ROCAUC of baseline capillary refill time was 0.73 [95% CI, 0.54-0.90] and of baseline PCO2gap was 0.79 [0.61-0.93]. Capillary refill time was significantly longer in non-survivors than in survivors at day 90.ConclusionΔCRT-PLR predicted peripheral perfusion response following volume expansion. This simple low-cost and non-invasive diagnostic method could be used in peripheral perfusion-targeted resuscitation protocols.Trial registrationCPP Lyon Sud-Est II ANSM: 2014-A01034-43 Clinicaltrial.gov, NCT02248025 , registered 13th of September 2014.

Dataset Information

Supervised Machine Learning Applied to Automate Flash and Prolonged Capillary Refill Detection by Pulse Oximetry.

Publications

Supervised Machine Learning Applied to Automate Flash and Prolonged Capillary Refill Detection by Pulse Oximetry.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets