Andreas Brink-Kjaer

Magnus Ruud Kjaer, Rahul Thapa, Gauri Ganjoo et al.

Polysomnography (PSG), the gold standard test for sleep analysis, generates vast amounts of multimodal clinical data, presenting an opportunity to leverage self-supervised representation learning (SSRL) for pre-training foundation models to enhance sleep analysis. However, progress in sleep foundation models is hindered by two key limitations: (1) the lack of a shared dataset and benchmark with diverse tasks for training and evaluation, and (2) the absence of a systematic evaluation of SSRL approaches across sleep-related tasks. To address these gaps, we introduce Stanford Sleep Bench, a large-scale PSG dataset comprising 17,467 recordings totaling over 163,000 hours from a major sleep clinic, including 13 clinical disease prediction tasks alongside canonical sleep-related tasks such as sleep staging, apnea diagnosis, and age estimation. We systematically evaluate SSRL pre-training methods on Stanford Sleep Bench, assessing downstream performance across four tasks: sleep staging, apnea diagnosis, age estimation, and disease and mortality prediction. Our results show that multiple pretraining methods achieve comparable performance for sleep staging, apnea diagnosis, and age estimation. However, for mortality and disease prediction, contrastive learning significantly outperforms other approaches while also converging faster during pretraining. To facilitate reproducibility and advance sleep research, we will release Stanford Sleep Bench along with pretrained model weights, training pipelines, and evaluation code.

William Lehn-Schiøler, Magnus Ruud Kjær, Phillip Hempel et al.

Sleep foundation models have recently demonstrated strong performance on in-domain polysomnography tasks, including sleep staging, apnea detection, and disease risk prediction. In this work, we investigate whether sleep biosignals can serve as an effective pretraining distribution for learning representations that transfer beyond sleep to adjacent domains. Following sleep foundation models, we perform sleep-only multimodal contrastive pretraining (with a leave-one-out objective) and evaluate transfer to non-sleep EEG and ECG, two well-benchmarked biosignal modalities with heterogeneous datasets and clinically meaningful downstream tasks. Across eight downstream tasks spanning multiple EEG and ECG datasets, sleep pretraining consistently improves performance relative to training from scratch. Moreover, on several tasks, we achieve performance competitive with or surpassing prior specialized state-of-the-art and foundation models.

Niels R. Lorenzen, Poul J. Jennum, Emmanuel Mignot et al.

Wearable accelerometers are widely used for continuous monitoring of physical activity. Supervised machine learning and deep learning algorithms have long been used to extract meaningful activity information from raw accelerometry data, but progress has been hampered by the limited amount of labeled data that is publicly available. Exploiting large unlabeled datasets using self-supervised pretraining is a relatively new and underexplored approach in the field of human activity recognition (HAR). We used a time-series transformer masked autoencoder (MAE) approach to self-supervised pretraining and propose two novel spectrogram-based loss functions: the log-scale meanmagnitude (LMM) and log-scale magnitude variance (LMV) losses. We compared these losses with the mean squared error (MSE) loss for MAE training. We leveraged the large unlabeled UK Biobank accelerometry dataset (n = 109k) for pretraining and evaluated downstream HAR performance using a linear classifier in a smaller labelled dataset. We found that pretraining with the LMM loss improved performance compared to an MAE pretrained with the MSE loss, with 12.7% increase in subject-wise F1 score when using linear probing. Compared with a state-of-the-art ResNet-based HAR model, our LMM-pretrained transformer models performed better (+9.8% F1) with linear probing and comparably when fine-tuned using an LSTM classifier. The addition of the LMV to the LMM loss decreased performance compared to the LMM loss alone. These findings establish the LMM loss as a robust and effective method for pretraining MAE models on accelerometer data for HAR and show the potential of pretraining sequence-based models for free-living HAR.

Andreas Brink-Kjaer, Alexander Neergaard Olesen, Paul E. Peppard et al.

Cortical arousals are transient events of disturbed sleep that occur spontaneously or in response to stimuli such as apneic events. The gold standard for arousal detection in human polysomnographic recordings (PSGs) is manual annotation by expert human scorers, a method with significant interscorer variability. In this study, we developed an automated method, the Multimodal Arousal Detector (MAD), to detect arousals using deep learning methods. The MAD was trained on 2,889 PSGs to detect both cortical arousals and wakefulness in 1 second intervals. Furthermore, the relationship between MAD-predicted labels on PSGs and next day mean sleep latency (MSL) on a multiple sleep latency test (MSLT), a reflection of daytime sleepiness, was analyzed in 1447 MSLT instances in 873 subjects. In a dataset of 1,026 PSGs, the MAD achieved a F1 score of 0.76 for arousal detection, while wakefulness was predicted with an accuracy of 0.95. In 60 PSGs scored by multiple human expert technicians, the MAD significantly outperformed the average human scorer for arousal detection with a difference in F1 score of 0.09. After controlling for other known covariates, a doubling of the arousal index was associated with an average decrease in MSL of 40 seconds ($β$ = -0.67, p = 0.0075). The MAD outperformed the average human expert and the MAD-predicted arousals were shown to be significant predictors of MSL, which demonstrate clinical validity the MAD.