Dataset · LANSG
ECG grounding with heartbeat feature alignment
LANSG/ECG-Grounding
30,000 instruction pairs for ECG interpretation with diagnosis justification based on specific physiological parameters.
What's inside
The dataset gathers 30,000 English instruction pairs for image-text-to-text tasks related to electrocardiogram analysis. Each entry is annotated at the individual heartbeat level with measurable physiological features on which the diagnosis is built.
What you can do with it
- Training multimodal medical models for ECG interpretation with reasoned conclusions
- Fine-tuning VLMs to improve the transparency and reliability of medical diagnostics
- Testing models for the ability to link ECG visual patterns with clinical signs
- Tasks
- Image text → text
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 427
- Likes
- 9
How to load it
from datasets import load_dataset
ds = load_dataset("LANSG/ECG-Grounding")