CaReSound
收藏资源简介:
CaReSound 是一个由荷兰埃因霍温理工大学研究人员创建的医疗音频数据集,包含呼吸和心脏声音记录,旨在推动诊断推理研究。该数据集包含来自 126 个主题的 6,898 个注释呼吸周期、112 个主题的呼吸声音记录、1,568 个主题的儿科心脏声音记录等,所有记录均附带详细元数据和配对问答示例。CaReSound 数据集通过整合自我监督音频编码器和大语言模型(LLMs)的推理能力,旨在为医疗诊断提供开放式的诊断答案,提高诊断准确性和深度理解潜在健康问题。
CaReSound is a medical audio dataset created by researchers at Eindhoven University of Technology in the Netherlands. It contains respiratory and cardiac sound recordings, aiming to promote diagnostic reasoning research. The dataset includes 6,898 annotated respiratory cycles from 126 subjects, respiratory sound recordings from 112 subjects, pediatric cardiac sound recordings from 1,568 subjects, and more. All recordings are accompanied by detailed metadata and paired question-answer examples. By integrating the inference capabilities of self-supervised audio encoders and large language models (LLMs), the CaReSound dataset is designed to provide open-ended diagnostic responses for medical diagnosis, improving diagnostic accuracy and enabling deeper understanding of underlying health conditions.

- 1CaReAQA: A Cardiac and Respiratory Audio Question Answering Model for Open-Ended Diagnostic Reasoning荷兰埃因霍温理工大学 · 2025年



