Dados do Trabalho

Título

Sleep prediction using data from oximeter, accelerometer and snoring from Biologix sleep study and artificial neural network

Introdução

Portable monitoring (PM) is a simplified and validated method for obstructive sleep apnea (OSA) diagnosis. However, PM does not usually detect sleep; the number of events is reported by hours of monitoring rather than hours of sleep. Therefore, the absence of sleep monitoring is a potential source of variability between polysomnography (PSG) and PM.

Objetivo

To develop and validate an artificial neural network (ANN) algorithm to predict sleep using data from the Biologix system, a new PM device.

Métodos

Patients with suspected OSA referred for PSG were studied with Biologix using the Oxistar™, Biologix Sistemas S.A., Brazil. The PSG and Biologix data were time-synchronized. To develop and validate the ANN algorithm for predicting sleep, we used data derived from the Biologix system, including oximeter (arterial oxygen saturation, heart rate), with built-in accelerometer (movement), and smartphone application (snoring). A k-fold cross-validation method (k=10) was applied, ensuring that all sleep studies were included in both training and testing across all iterations. Sensitivity, specificity, and accuracy were calculated to evaluate the performance of the ANN algorithm. To assess the agreement on OSA diagnosis between PSG-apnea-hypopnea index (AHI) and Biologix-oxygen desaturation index (ODI), without and with sleep prediction, Bland-Altman plots were performed.

Resultados

A total of 268 patients were studied (age: 56±11 years; body mass index: 30.9±4.6 kg/m², AHI: 35±30 events/h). The sleep studies underwent 10-fold cross-validation. Each fold consisted of a training set of 90% of the patients (approximately 241 patients, corresponding to 221,639 epochs) and a test set of 10% of the patients (approximately 27 patients, corresponding to 24,627 epochs). The final ANN algorithm exhibited a sensitivity of 91.5%, specificity of 71.0%, and accuracy of 86.1% to detect sleep. Compared to the Biologix-ODI without sleep prediction, the bias (mean difference) between PSG-AHI and Biologix-ODI with sleep prediction decreased significantly (3.40 vs. 1.02 events/h, p<0.001).

Conclusões

The ANN model using data from the Biologix system, including oximeter, accelerometer, and snoring, is able to detect sleep with good accuracy.

Palavras -chave

Sleep apnea, obstructive; oximetry; actigraphy; snoring; model, neural network.