Research map / Time series / Foundation models for time series
Sequential Multi-Dimensional Self-Supervised Learning for Clinical Time Series
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add component-level SSL loss to global SSL (two-level SMD SSL) with SimCLR | SimCLR trajectory pre-training · Dataset 1 and 2 unimodal | AUROC (Elevated mPAP, 24hr Mortality) | Helped |
| Add component-level loss to global VICReg (two-level SMD SSL) | VICReg pre-training · Dataset 1 and 2 unimodal | AUROC | Mixed |
| Pre-training with SMD SSL vs no pre-training | SMD SSL (SimCLR/VICReg) · Dataset 2 multimodal, 24hr Mortality | AUROC | Mixed |
| Add multimodal (structured + signals) input vs signals-only | Pre-training with SMD SSL · Dataset 2 Elevated mPAP | AUROC: 71.6 multimodal vs 67.0 signals-only (SMD SSL VICReg vs SimCLR unimodal) | Helped |
| Pre-training longer (50 vs 15 epochs) for SMD SSL (VICReg) (ablation) | SMD SSL (VICReg) · Dataset 2 multimodal, Elevated mPAP and 24hr Mortality | AUROC: 71.6 to 66.6 (mPAP); 90.7 to 89.3 (mortality) | Hurt |
| Full fine-tuning instead of linear evaluation (ablation) | SMD SSL (VICReg) and baselines · Dataset 2 multimodal, 24hr Mortality | AUROC: 90.7 full FT vs 72.2 linear evaluation | Helped |
| Add structured-only pre-training (SSL on structured) vs RandInit (ablation) | Structured-only inputs · Dataset 1 and 2 multimodal, 24hr Mortality | AUROC: 79.0 vs 79.1 (Dataset 1); 89.3 vs 90.0 (Dataset 2) | No clear effect |
Closest papers by meaning
- Integrating Sequence and Image Modeling in Irregular Medical Time Series Through Self-Supervised Learning (AAAI 2025 · similarity 0.89)
- From Token to Rhythm: A Multi-Scale Approach for ECG-Language Pretraining (ICML 2025 · similarity 0.87)
- Self-Supervised Framework Based on Subject-Wise Clustering for Human Subject Time Series Data (AAAI 2024 · similarity 0.86)
- Reading Your Heart: Learning ECG Words and Sentences via Pre-training ECG Language Model (ICLR 2025 · similarity 0.86)
- Parallel-Learning of Invariant and Tempo-variant Attributes of Single-Lead Cardiac Signals: PLITA (AAAI 2025 · similarity 0.85)
- TimeMIL: Advancing Multivariate Time Series Classification via a Time-aware Multiple Instance Learning (ICML 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.