Seismocardiography Pig Hypovolemia Dataset for Signal Quality Indexing and Validated Cardiac Timings

Description Seismocardiography (SCG), a non-invasive method for capturing cardio-mechanical signals, shows promise for evaluating cardiovascular health. However, SCG is susceptible to noise and motion artifacts, complicating accurate signal quality indexing and fiducial point detection. To address this, we present a dataset of expert-annotated SCG waveforms from a porcine hypovolemia protocol. This dataset contains signal quality scores and annotated cardiac timing intervals for 17,059 SCG heartbeats across five porcine subjects. The annotations were created using a custom-developed open-source graphical user interface. Key annotated fiducial points include aortic opening (AO) and aortic closing (AC), which were validated against gold-standard invasive catheter-based measurements, achieving strong correlations (r=0.926 for AO and r=0.911 for AC). The dataset is intended to facilitate the standardized evaluation of signal quality and cardiac fiducial point detection algorithms. It provides a foundational resource for the development and benchmarking of machine learning models in cardio-mechanical signal processing, supporting advancements in real-time cardiac monitoring, denoising, and diagnostic applications, thereby enhancing reproducibility in SCG research. Dataset Raw. The Raw data package contains the original.mat data collected from the initial hypovolemia protocol. The data is organized with all pig subjects in one folder. The naming convention for each file is pig_[ID]_all.mat, where ID represents the subject ID for the corresponding pig subject. Signal descriptions, extracted features, and further information can be found in previous studies. Annotated. The Annotated data package contains individual annotations, median annotations, extracted AoP catheter timings for AO and AC, and filtered ECG, SCG, and AoP segmented waveforms used for annotations and analyses. These data are packaged in a common .csv format to fit a wide variety of user applications. Each pig subject contains their own folder in Annotated with a naming convention of pig_[ID], where ID represents the subject ID for the corresponding pig subject. The table below shows the list of files, file contents, and data points associated with each label. File Name Data Columns Data Value Value Interpretation annotation_points.csv [AO_1, AO_2, AO_3] AO Timings Millisecond [ms] [AC_1, AC_2, AC_3] AC Timings Millisecond [ms] [ACv_1, ACv_2, ACv_3] AC Valley Timings Millisecond [ms] [MO_1, MO_2, MO_3] MO Timings Millisecond [ms] [SQI_1, SQI_2, SQI_3] SQI Ratings ‘Good’=2, Average’=1, Bad’=0 [shifted_1, shifted_2, shifted_3] Shifted Beat Flag ‘True’=1, False’=0 annotation_points_median.csv [AO] AO Timings Millisecond [ms] [AC] AC Timings Millisecond [ms] [ACv] AC Valley Timings Millisecond [ms] [MO] MO Timings Millisecond [ms] [SQI] SQI Ratings ‘Good’=2, Average’=1, Bad’=0 [shifted] Shifted Beat Flag ‘True’=1, False’=0 cath_points.csv [AO] AO Timings Millisecond [ms] [AC] AC Timings Millisecond [ms] aop.csv [1,2, 3…, n] Amplitude of Sampling Frequency Index Millimeters of Mercury [mmHg] ecg.csv [1,2, 3…, n] Amplitude of Sampling Frequency Index Millivolt [mv] scg.csv [1,2, 3…, n] Amplitude of Sampling Frequency Index Acceleration [mg]