Effects of precise cardio sounds on the success rate of phonocardiography

Code and Dataset for 'Effects of precise cardio sounds on the success rate of phonocardiography' on PLOS ONE<br>The protocol for this study was approved by the Institutional Review Board for Seoul St.Mary’s Hospital (approval no. KC17TESI0684 ).<br>Infrasound_Dataset is deep learning model which train &amp; validate from our infrasound datasets with Tensorflow version 1.14dataset file is .hdf5 files which can load with python.The structure of the dataset file is shown below.normal data: 0, 1, sex_data, age_data, fold_label, flatten data(2d CWT)abnormal data: 1, 0, sex_data, age_data, fold_label, flatten data(2d CWT)Reference Files is a list of labels for datasets, separated by measurement period.Infrasound dataset's Deep learning model &amp; train file is classify_cnn_keras_noReg_crb.py &amp; classify_lstm_keras_noReg_crb.pyPhysioNet_Dataset is deep learning model which train &amp; validate from PhysioNet CinC 2016 datasets with Tensorflow version 1.14PhysioNet's Deep learning model &amp; train file is classify_cnn_keras_noreg_crb.py &amp; classify_lstm_keras_noreg_crb.pyThe code_heartsound.m file performs a Continuous Wavelet Transform (CWT) on an audio file and saves the results in a CSV fileFig2 is the code for draw FFT results to compare Normal &amp; Abnormal dataFig3 is the code for draw Time-domain &amp; CWT &amp; STFT data using one example dataThe original audio files without any pre-processing can be requested by contacting the authors Mi-Hyung Moon([e-mail]) or Young-Sin Kim([e-mail])发表于《PLOS ONE》的论文《精准心音对心音图检查成功率的影响》配套代码与数据集。本研究方案已获首尔圣母医院机构审查委员会批准（批准号：KC17TESI0684）。 Infrasound_Dataset是基于自有次声数据集，使用TensorFlow 1.14版本进行训练与验证的深度学习模型。数据集文件格式为.hdf5，可通过Python加载。数据集文件结构如下：正常样本：0, 1, 性别数据、年龄数据、折次标签、扁平化二维连续小波变换（Continuous Wavelet Transform, CWT）数据；异常样本：1, 0, 性别数据、年龄数据、折次标签、扁平化二维连续小波变换数据。参考文件为按测量周期划分的数据集标签列表。 次声数据集对应的深度学习模型及训练脚本为classify_cnn_keras_noReg_crb.py与classify_lstm_keras_noReg_crb.py。 PhysioNet_Dataset是基于PhysioNet CinC 2016数据集，使用TensorFlow 1.14版本进行训练与验证的深度学习模型。PhysioNet数据集对应的深度学习模型及训练脚本为classify_cnn_keras_noreg_crb.py与classify_lstm_keras_noreg_crb.py。 code_heartsound.m脚本可对音频文件执行连续小波变换（Continuous Wavelet Transform, CWT），并将变换结果保存为CSV格式文件。 Fig2为用于绘制快速傅里叶变换（Fast Fourier Transform, FFT）结果以对比正常与异常样本的代码。 Fig3为使用单条示例数据绘制时域、连续小波变换（CWT）及短时傅里叶变换（Short-Time Fourier Transform, STFT）结果的代码。 如需获取未经过任何预处理的原始音频文件，可联系作者文美香（Mi-Hyung Moon，邮箱：[e-mail]）或金荣信（Young-Sin Kim，邮箱：[e-mail]）。