all AI news
Zero-Shot ECG Classification with Multimodal Learning and Test-time Clinical Knowledge Enhancement
March 12, 2024, 4:43 a.m. | Che Liu, Zhongwei Wan, Cheng Ouyang, Anand Shah, Wenjia Bai, Rossella Arcucci
cs.LG updates on arXiv.org arxiv.org
Abstract: Electrocardiograms (ECGs) are non-invasive diagnostic tools crucial for detecting cardiac arrhythmic diseases in clinical practice. While ECG Self-supervised Learning (eSSL) methods show promise in representation learning from unannotated ECG data, they often overlook the clinical knowledge that can be found in reports. This oversight and the requirement for annotated samples for downstream tasks limit eSSL's versatility. In this work, we address these issues with the Multimodal ECG Representation Learning (MERL}) framework. Through multimodal learning on …
abstract arxiv classification clinical cs.ai cs.lg data diagnostic diseases eess.sp found knowledge multimodal multimodal learning oversight practice reports representation representation learning self-supervised learning show supervised learning test tools type zero-shot
More from arxiv.org / cs.LG updates on arXiv.org
Jobs in AI, ML, Big Data
Software Engineer for AI Training Data (School Specific)
@ G2i Inc | Remote
Software Engineer for AI Training Data (Python)
@ G2i Inc | Remote
Software Engineer for AI Training Data (Tier 2)
@ G2i Inc | Remote
Data Engineer
@ Lemon.io | Remote: Europe, LATAM, Canada, UK, Asia, Oceania
Artificial Intelligence – Bioinformatic Expert
@ University of Texas Medical Branch | Galveston, TX
Lead Developer (AI)
@ Cere Network | San Francisco, US