William Lehn-Schiøler

Anders Gjølbye, William Lehn-Schiøler, Áshildur Jónsdóttir et al.

Deep learning models are complex due to their size, structure, and inherent randomness in training procedures. Additional complexity arises from the selection of datasets and inductive biases. Addressing these challenges for explainability, Kim et al. (2018) introduced Concept Activation Vectors (CAVs), which aim to understand deep models' internal states in terms of human-aligned concepts. These concepts correspond to directions in latent space, identified using linear discriminants. Although this method was first applied to image classification, it was later adapted to other domains, including natural language processing. In this work, we attempt to apply the method to electroencephalogram (EEG) data for explainability in Kostas et al.'s BENDR (2021), a large-scale transformer model. A crucial part of this endeavor involves defining the explanatory concepts and selecting relevant datasets to ground concepts in the latent space. Our focus is on two mechanisms for EEG concept formation: the use of externally labeled EEG datasets, and the application of anatomically defined concepts. The former approach is a straightforward generalization of methods used in image classification, while the latter is novel and specific to EEG. We present evidence that both approaches to concept formation yield valuable insights into the representations learned by deep EEG models.

Anders Gjølbye, Lina Skerath, William Lehn-Schiøler et al.

Electroencephalography (EEG) research typically focuses on tasks with narrowly defined objectives, but recent studies are expanding into the use of unlabeled data within larger models, aiming for a broader range of applications. This addresses a critical challenge in EEG research. For example, Kostas et al. (2021) show that self-supervised learning (SSL) outperforms traditional supervised methods. Given the high noise levels in EEG data, we argue that further improvements are possible with additional preprocessing. Current preprocessing methods often fail to efficiently manage the large data volumes required for SSL, due to their lack of optimization, reliance on subjective manual corrections, and validation processes or inflexible protocols that limit SSL. We propose a Python-based EEG preprocessing pipeline optimized for self-supervised learning, designed to efficiently process large-scale data. This optimization not only stabilizes self-supervised training but also enhances performance on downstream tasks compared to training with raw data.

William Lehn-Schiøler, Magnus Ruud Kjær, Rahul Thapa et al.

EEG foundation models achieve state-of-the-art clinical performance, yet the internal computations driving their predictions remain opaque: a barrier to clinical trust. We apply TopK Sparse Autoencoders (SAEs) across three architecturally distinct EEG transformers: SleepFM, REVE, and LaBraM to extract sparse feature dictionaries from their embeddings. By grounding these features in a clinical taxonomy (abnormality, age, sex, and medication), we benchmark monosemanticity and entanglement across architectures. A single hyperparameter procedure, driven by an intrinsic dictionary health audit, transfers robustly across all three architectures. Via concept steering, we introduce a "target vs. off-target" probe area metric to quantify steering selectivity and reveal three operational regimes: selectively steerable, encoded but entangled, and non-encoded. This framework exposes critical representational failures: "wrecking-ball" interventions that collapse global model performance, and clinical entanglements, such as age-pathology confounding, where it is impossible to suppress one concept without corrupting the other. Finally, a spectral decoder maps these interventions back to the amplitude spectrum, translating latent manipulations into physiologically interpretable frequency signatures, such as pathological slow-wave suppression and $α$-band restoration.

Nomin Enkhtsetseg, William Lehn-Schiøler, Anton Storgaard Mosquera et al.

Purpose: Access to electroencephalography (EEG) remains limited across low- and middle-income countries (LMICs) due to cost, infrastructure requirements, and a shortage of trained staff. This study evaluated the feasibility and clinical utility of a smartphone-based EEG system in a real-world setting. Methods: We conducted a multicenter observational study (November 2023 to April 2026) across 29 clinical sites in Kenya. A smartphone-based 27-lead EEG system enabled trained healthcare workers to acquire standardized recordings with remote expert interpretation. Results: 3,036 EEG sessions were performed. Male patients constituted 57.8% of the cohort, with representation across pediatric and adult populations. The most common referral indication was seizures or convulsions (68.5%). Overall, 2,915 (96%) recordings were interpretable, while 121 (4%) were uninterpretable, primarily due to high electrode impedance and insufficient recording duration. Uninterpretable recordings were significantly shorter than interpretable recordings (mean 18.5 vs. 33.8 minutes; median 15.1 vs. 31.6 minutes; p < 0.0001). Mean turnaround time for interpretation was 107 minutes. Among interpretable recordings, 917 (30.2%) were abnormal, including 701 (76.4%) with epileptiform abnormalities, 215 (23.4%) with non-epileptiform findings, and 1 (0.1%) indeterminate finding. Epileptiform abnormalities were highest in children aged 4-9 years (33.1%) and less frequent in adults (14-21%). Non-epileptiform abnormalities were more common in patients aged 60+ years (19.2%) compared to younger age groups (3-9%). Conclusion: Large-scale, point-of-care EEG acquisition by non-specialist operators in a resource-limited setting is feasible. Expansion of smartphone-based EEG systems may improve equitable access to neurological diagnosis and care in LMICs.

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.