Online Retraining

Select a model.
Select a previously trained model.
Train a new predictive model for Sleep Staging or Sleep-Disordered Breathing. While training is in progress, you may navigate away from the page or close your browser. Trained models may be saved for later use in the Analysis section of the platform. Select an existing model to test on out-of-sample data, or configure and train a new model.

Sleep-Disordered Breathing

Model configuration.
Configuration
Select between Sequential Minimal Optimization (SMO) or Iterative Single Data Algorithm (ISDA). ISDA uses a rho (ρ)-free formulation and is not available for One-Class SVM. SMO is recommended for most use-cases. Select a binary classifier to predict whether the apnea-hypopnea index (AHI) of a child is ≥5 (moderate-severe sleep disordered breathing) or to classify a segment of recording as Sleep or Wake. Alternatively, you may select a regressor to produce a point estimate of the AHI and estimates of uncertainty. Select a Linear, Gaussian, or Polynomial kernel. The Linear kernel enables better interpretability, while the Gaussian kernel is more flexible and may capture non-linearities. Polynomial kernels offer a balance in complexity.
Model Information
Feature Selection
Feature Description
Desats Number of 3% drops (desaturations) in the SpO₂ per hour.
Desat Depth Cumulative desaturation depth per hour.
Desat Duration Cumulative desaturation duration per hour.
Desat Area Cumulative desaturation area per hour.
D90 Desaturations below 90% SpO₂ per hour.
D85 Desaturations below 85% SpO₂ per hour.
T95 Fraction of recording below 95% SpO₂.
T90 Fraction of recording below 90% SpO₂.
OSA Band Relative band power of the SpO₂ in the OSA Band (0.02-0.044Hz).
OSA Band Power Mean Mean frequency (spectral centroid) of the OSA Band Welch-averaged band power spectrum.
OSA Band Power Variance Second central moment about frequency (spectral variance) of the OSA Band Welch-averaged band power spectrum.
OSA Band Power Skewness Third standardized central moment about frequency (spectral skewness) of the OSA Band Welch-averaged band power spectrum.
OSA Band Power Kurtosis Fourth standardized central moment about frequency (spectral excess kurtosis) of the OSA Band Welch-averaged band power spectrum.
OSA Band Power Peak Frequency Frequency of the dominant peak within the OSA Band Welch-averaged band power spectrum.
OSA Band Power Peak Area Summed power of the dominant peak within the OSA Band Welch-averaged band power spectrum.
OSA Band Spectral Entropy Shannon entropy of the normalized OSA Band Welch-averaged band power spectrum.
OSA Band Spectral Entropy² Shannon entropy of the normalized squared (second order) OSA Band Welch-averaged band power spectrum.
OSA Band Spectral Entropy³ Shannon entropy of the normalized cubed (third order) OSA Band Welch-averaged band power spectrum.
SpO₂ Mean Mean of the SpO₂.
SpO₂ Variance Second central moment of the SpO₂.
SpO₂ Skewness Third standardized central moment of the SpO₂.
SpO₂ Kurtosis Fourth standardized central moment (excess kurtosis) of the SpO₂.
SpO₂ Complexity Lempel-Ziv Complexity of the binarized SpO₂. A measure of repeating sequences in the SpO₂.
SpO₂ Shannon Entropy Shannon entropy of the SpO₂. A static measure of regularity.
SpO₂ Permutation Entropy Permutation entropy of the SpO₂ where the embedding delay equals 1, and m equals 3. A dynamic measure of regularity.
SpO₂ Sample Entropy Sample entropy of the SpO₂ where the template length equals 2, and the tolerance equals the product of 0.2 and the standard deviation of the SpO₂. A dynamic measure of regularity.
SpO₂ SD1 Standard deviation of the first difference in the SpO₂. A measure of short-term SpO₂ variability.
SpO₂ SD2 Long-axis spread of the SpO₂ Poincaré plot, reflecting long-term SpO₂ variability.
SpO₂ SD1/SD2 Ratio of short-term and long-term variability in the SpO₂.
SpO₂ Power Total Welch-averaged band power (%²) of the SpO₂ in the 0.01-0.25Hz range.
SpO₂ Power Mean Mean frequency (spectral centroid) of the SpO₂ Welch-averaged band power spectrum.
SpO₂ Power Variance Second central moment about frequency (spectral variance) of the SpO₂ Welch-averaged band power spectrum.
SpO₂ Power Skewness Third standardized central moment about frequency (spectral skewness) of the SpO₂ Welch-averaged band power spectrum.
SpO₂ Power Kurtosis Fourth standardized central moment about frequency (spectral excess kurtosis) of the SpO₂ Welch-averaged band power spectrum.
SpO₂ Power Peak Frequency Frequency of the dominant peak within the SpO₂ Welch-averaged band power spectrum.
SpO₂ Power Peak Area Summed power of the dominant peak within the SpO₂ Welch-averaged band power spectrum.
SpO₂ Spectral Entropy Shannon entropy of the normalized SpO₂ Welch-averaged band power spectrum.
SpO₂ Spectral Entropy² Shannon entropy of the normalized squared (second order) SpO₂ Welch-averaged band power spectrum.
SpO₂ Spectral Entropy³ Shannon entropy of the normalized cubed (third order) SpO₂ Welch-averaged band power spectrum.
SpO₂ DFA Short-range detrended fluctuation analysis of the SpO₂ using 10 log-spaced windows from 20 to 200 seconds. Captures short-timescale self-affinity.
Pulse Rate Rise Index 10 Number of rises in the Pulse Rate of at least 10 beats per minute, per hour.
Pulse Rate Rise Index 12 Number of rises in the Pulse Rate of at least 12 beats per minute, per hour.
Pulse Rate Rise Index 15 Number of rises in the Pulse Rate of at least 15 beats per minute, per hour.
Pulse Rate VLF Band Relative band power of the Pulse Rate in the VLF Band (0.01-0.04Hz).
Pulse Rate LF Band Relative band power of the Pulse Rate in the LF Band (0.04-0.15Hz).
Pulse Rate HF Band Relative band power of the Pulse Rate in the HF Band (0.15Hz to 0.25Hz).
Pulse Rate LF/HF Ratio of band powers in the LF and HF bands of the Pulse Rate. An estimate of sympathovagal balance.
Pulse Rate Mean Mean of the Pulse Rate.
Pulse Rate Variance Second central moment of the Pulse Rate.
Pulse Rate Skewness Third standardized central moment of the Pulse Rate.
Pulse Rate Kurtosis Fourth standardized central (excess kurtosis) moment of the Pulse Rate.
Pulse Rate Complexity Lempel-Ziv Complexity of the binarized Pulse Rate. A measure of repeating sequences in the Pulse Rate.
Pulse Rate Shannon Entropy Shannon entropy of the Pulse Rate. A static measure of regularity.
Pulse Rate Permutation Entropy Permutation entropy of the Pulse Rate where the embedding delay equals 1, and m equals 3. A dynamic measure of regularity.
Pulse Rate Sample Entropy Sample entropy of the Pulse Rate where the template length equals 2, and the tolerance equals the product of 0.2 and the standard deviation of the Pulse Rate. A dynamic measure of regularity.
Pulse Rate SD1 Standard deviation of the first difference in the Pulse Rate. A measure of short-term Pulse Rate variability.
Pulse Rate SD2 Long-axis spread of the Pulse Rate Poincaré plot, reflecting long-term Pulse Rate variability.
Pulse Rate SD1/SD2 Ratio of short-term and long-term variability in the Pulse Rate.
Pulse Rate Power Total Welch-averaged band power (bpm²) of the Pulse Rate in the 0.01-0.25Hz range.
Pulse Rate Power Mean Mean frequency (spectral centroid) of the Pulse Rate Welch-averaged band power spectrum.
Pulse Rate Power Variance Second central moment about frequency (spectral variance) of the Pulse Rate Welch-averaged band power spectrum.
Pulse Rate Power Skewness Third standardized central moment about frequency (spectral skewness) of the Pulse Rate Welch-averaged band power spectrum.
Pulse Rate Power Kurtosis Fourth standardized central moment about frequency (spectral excess kurtosis) of the Pulse Rate Welch-averaged band power spectrum.
Pulse Rate Power Peak Frequency Frequency of the dominant peak within the Pulse Rate Welch-averaged band power spectrum.
Pulse Rate Power Peak Area Summed power of the dominant peak within the Pulse Rate Welch-averaged band power spectrum.
Pulse Rate Spectral Entropy Shannon entropy of the normalized Pulse Rate Welch-averaged band power spectrum.
Pulse Rate Spectral Entropy² Shannon entropy of the normalized squared (second order) Pulse Rate Welch-averaged band power spectrum.
Pulse Rate Spectral Entropy³ Shannon entropy of the normalized cubed (third order) Pulse Rate Welch-averaged band power spectrum.
Pulse Rate DFA Short-range detrended fluctuation analysis of the Pulse Rate using 10 log-spaced windows from 20 to 200 seconds. Captures short-timescale self-affinity.
Select training dataset.
Select a precomputed dataset to use for model training. Only datasets specifically configured for training are available. Cross-validation results are available during model testing.
Training configuration and summary.
Bayesian Optimization
Number of objective function evaluations.
Number of candidates evaluated per iteration.
Configure the Bayesian Optimizer and search space for each hyperparameter. The number of iterations determines how many points are evaluated, and the number of candidates controls the resolution of the acquisition search.
Training Progress
Training in progress. You may use the analysis functionality or close your browser.

Optimization Iteration 14 of 30

Current Parameters: C=938.438, Epsilon=2.407, Gamma=0.002
0
5
10
15
20
25
30
35
40
45
50
55
60
65
70
Iteration
Mean Squared Error (MSE)
Bayesian Optimization is used to search the hyperparameter ranges provided. This process attempts to balance exploration and exploitation to select the optimal values for the trained model. Gaussian Process Regression (GPR) and the Expected Improvement acquisition function are used. Model Validation data are collected from out-of-fold testing during cross-validation. These data are saved with the model and do not change. Models are calibrated using cross-validation data when training is complete so that predictions are accompanied by estimates of uncertainty. A trained Ridge Logistic Regression model is embedded within Support Vector Classifiers to produce a probability point-estimate during prediction, and a Laplace distribution is computed and embedded within Linear and Support Vector Regression to produce confidence intervals that accompany the AHI point-estimate during prediction. Trained models may be saved and are immediately accessible through the Analysis and Batch Analysis areas of the platform.