CentificAIResearch/MedSpectrogram-Reasoning
收藏资源简介:
MedSpectrogram-Reasoning是一个医疗声学问答数据集,专门用于基于梅尔频谱图(mel spectrograms)进行推理,涵盖心音、肺音和咳嗽声。每个记录包含一个临床问题、录音的梅尔频谱图像、真实答案以及录音的临床元数据。该任务为纯视觉任务:模型接收频谱图像和问题,必须通过频谱图推理生成答案,原始音频不作为模型输入,仅用于参考和可重复性。本版本为分层样本,覆盖所有声音类型和源数据集。测试集包括126个录音和504个问答对,来自多个公开医疗声音数据集,如CirCor DigiScope(心音)、ICBHI 2017(肺音)、Coswara(咳嗽)等。数据以Parquet和JSONL格式存储,包含音频文件(.wav)和频谱图图像(.png)。数据集用于研究和教育目的,采用CC-BY-4.0许可证。
MedSpectrogram-Reasoning is a medical acoustic question-answering dataset for reasoning over mel spectrograms of heart, lung, and cough sounds. Each record provides a clinical question, the mel-spectrogram image of a recording, a ground-truth answer, and the recordings clinical metadata. The task is purely visual: a model receives the spectrogram image together with the question and must reason over the spectrogram to produce the answer. The raw audio is not used as model input—the original .wav recordings are included only for reference and reproducibility. This release is a stratified sample covering all sound types and source datasets. The test split includes 126 recordings and 504 QA pairs, sourced from publicly available medical sound datasets such as CirCor DigiScope (heart), ICBHI 2017 (lung), and Coswara (cough). Data is stored in Parquet and JSONL formats, with audio files (.wav) and spectrogram images (.png). The dataset is intended for research and educational purposes, released under the CC-BY-4.0 license.




