Research map / Label-efficient and robust learning / Recognition and long-tail
CROCS: Clustering and Retrieval of Cardiac Signals Based on Patient Disease Class, Sex, and Age
Dani Kiyasseh, Tingting Zhu, David A. Clifton
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Use clinical prototypes as centroids rather than traditional prototypes | CROCS clustering on Chapman · Chapman validation | class accuracy: 80.3 -> 90.3 | Improved |
| Use clinical prototypes rather than DTC prototypes for retrieval | CROCS retrieval on Chapman · Chapman validation, K=1, at least 1 attribute match | precision: 71.9 -> 95.6 | Improved |
| Use soft rather than hard assignment (ablation) | CROCS clustering · PTB-XL validation | class accuracy: 66.5 -> 76.0 | Improved |
| Use modulated rather than uniform soft attraction (ablation) | CROCS clustering · Chapman validation | class AMI: 68.2 -> 72.1 | Improved |
| Add the semantic-distance regularizer (ablation) | CROCS clustering · Chapman validation | class AMI: 72.1 -> 72.8 | Improved |
| Set β to 0.2 rather than 0.05, 0.1, or 0.4 (ablation) | CROCS clustering · PTB-XL validation | class accuracy: 66.0 -> 76.0; 65.7 -> 76.0; 65.9 -> 76.0 | Improved |
| Use CP CROCS rather than TP CROCS as the retrieval query | CROCS retrieval · PTB-XL validation, K=1, at least 1 attribute match | precision: 99.0 -> 92.5 | Hurt |
| Reduce labelled training data to 10% | CROCS clustering · Chapman validation | class accuracy: 90.3 -> 82.7 | Hurt |
| Vary embedding dimension across the tested sizes (ablation) | CROCS clustering · Chapman class and sex assignment | accuracy | No clear effect |
| Use β=0.2 instead of other tested β values (ablation) | CROCS retrieval · PTB-XL validation, K=1, at least 1 attribute match | precision: 63.5 -> 68.8; 63.5 -> 68.8; 67.5 -> 68.8 | Improved |
Closest papers by meaning
- Boosting Masked ECG-Text Auto-Encoders as Discriminative Learners (ICML 2025 · similarity 0.83)
- CardioLens: Revealing the Clinical Reality Gap of MLLMs via Multi-Sequence Cardiac MRI Evaluations (NeurIPS 2026 · similarity 0.83)
- Contrastive Learning for Clinical Outcome Prediction with Partial Data Sources (ICML 2024 · similarity 0.83)
- Self-supervised Learning of Echocardiographic Video Representations via Online Cluster Distillation (NeurIPS 2025 · similarity 0.83)
- CMR-RD: Long-Tailed Adaptive VLM for Explainable CMR Diagnosis (CVPR 2026 · similarity 0.82)
- GRAPE: Graph-Augmented Prototype Explanations for Interactive Medical Image Diagnosis (NeurIPS 2026 · similarity 0.82)
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.
