This service uses machine learning (ML) to analyse polysomnography (PSG) data in children. Computed statistical, non-linear, and spectral features are used in the prediction process.

Peer-reviewed article containing out-of-sample performance data coming soon.

Sleep Staging

Sleep staging uses multiclass classification and the Softmax function. Thirty-second segments (epochs) of the recording are classified as Stage Wake, N1, N2, N3, or REM. Log probability smoothing is then used to stabilise scored stages and transitions. These data are used to compute sleep statistics traditionally obtained from overnight polysomnography, such as sleep duration, latency, efficiency, and WASO.

Respiratory Scoring

Respiratory scoring uses multiclass classification and the Softmax function. Six-second segments (epochs) of the recording are assigned a classification of No Event, Obstructive Apnoea, Central Apnoea, and Hypopnea (considered an obstructive event) as appropriate. Log probability smoothing and a heuristic layer are then used to stabilise scored respiratory events. These data are used to compute respiratory event statistics traditionally obtained from overnight polysomnography, such as the apnoea-hypopnea index (AHI), central AHI, and obstructive AHI.

Online Retraining

This service offers a streamlined pipeline for training and validating models using a bank of pre-computed features and datasets. Newly trained models are immediately available for use after saving, enabling rapid prototyping and rollout of updated models. All models produce a set of calibrated probabilities for each stage of sleep.

Model training uses a set of custom solvers. Gradient Boosting Machines (GBMs) are Gradient Boosting Decision Trees that use an XGBoost-like algorithm and are trained using the Histogram Tree Method. Multinomial Regression models are trained using the Limited-memory BFGS (L-BFGS) Method. The performance of these solvers is comparable to other widely used machine learning libraries.