Deviations in sleep patterns during pregnancy, detected with machine-learning algorithms, can predict the risk of preterm birth, Stanford School of Medicine research found.