Update README.md

README.md

@@ -3,8 +3,8 @@ license: mit
 ---
 # FSL_ECG_QA_Dataset
  
-FSL_ECG_QA_Dataset is a benchmark dataset designed for open-ended diagnostic reasoning using cardiac and respiratory auscultation audio. It includes annotated medical sound recordings enriched with metadata and automatically generated question-answer (QA) pairs, enabling research into audio-language modeling for healthcare.
- 
+**FSL_ECG_QA_Dataset** is a **benchmark dataset** specifically designed to accompany the paper *"Electrocardiogram–Language Model for Few-Shot Question Answering with Meta Learning"* (**arXiv:2410.14464v1**). It supports research in combining **electrocardiogram (ECG) signals** with **natural language question answering (QA)**, particularly in **few-shot** and **meta-learning** scenarios.
+
 ## Dataset Highlights
  
 - Multimodal: Each sample pairs an auscultation audio clip with metadata and diagnostic question-answer pairs.
@@ -13,7 +13,7 @@ FSL_ECG_QA_Dataset is a benchmark dataset designed for open-ended diagnostic rea
  
 ## Source Datasets
  
-CaReSound is curated from several open-access medical audio datasets:
+The dataset is a structured reorganization of the existing ECG-QA dataset, adapted to suit meta-learning tasks. It draws samples from ECG sources such as PTB-XL and MIMIC-IV-ECG, and organizes them into diverse task sets based on question types (e.g., binary, multiple-choice, and query-based) and clinical attributes (e.g., SCP codes, noise type, axis deviation). This structure enables models to rapidly adapt to new diagnostic tasks with limited annotated examples.
  
 - **ICBHI Respiratory Sound Dataset**  
   download from the official ICBHI challenge website: [ICBHI 2017 Respiratory Sound Dataset](https://bhichallenge.med.auth.gr)  
@@ -32,6 +32,12 @@ CaReSound is curated from several open-access medical audio datasets:
 - **ZCHSound Pediatric Heart Sound Dataset**  
   download from [ZCHSound](http://zchsound.ncrcch.org.cn)
  
+
+To utilize this dataset, the authors propose a novel multimodal meta-learning framework that integrates a frozen ECG encoder, a frozen language model (e.g., LLaMA or Gemma), and a trainable cross-modal fusion module. This setup effectively aligns ECG signals with natural language queries to enable accurate clinical question answering.
+
+Experimental results demonstrate that under a 5-way 5-shot setting, the proposed method consistently outperforms conventional supervised baselines across different question types, showing strong generalization capabilities.
+
+In summary, FSL_ECG_QA_Dataset serves as a powerful benchmark for developing robust and generalizable ECG-based QA systems in data-scarce clinical environments. 
  
 ## QA Generation