 In this study, we proposed a novel approach to detect anomalous heart rate patterns in sleep time data collected using wearable devices. Our methodology consisted of a disentangled variational autoencoder, beta VAE, with a bi-LSTM backend, which was trained on data from eight different participants. We then tested the performance of our model against other well-known anomaly detection algorithms and found that it outperformed them in most cases. Additionally, we suggested that wearable devices could benefit from integrating anomaly detection algorithms as they would provide users with more processed and straightforward information. This article was authored by Alessio Staffani, Thomas Vincent, Anayal Chung, and others.