Smartphone-Based Seismocardiography Dataset: Raw Accelerometer Signals and Clinical Data from Healthy Subjects and Cardiac Patients
Description
This dataset contains raw seismocardiographic (SCG) signals and associated clinical data from a cross-sectional observational study comparing healthy individuals and cardiac patients. The study aimed to develop and validate a non-invasive method for acquiring SCG signals using smartphone accelerometers. Study Design: Cross-sectional observational study conducted between July 2025 and August 2026 at the University of Southern Santa Catarina, Brazil. The study was approved by the Research Ethics Committee (protocol number 8.033.743, CAAE 92999425.6.0000.0261). Participants: A total of 60 participants were included, divided into two groups: Healthy group (n = 30): Healthy adults recruited from the community, with no diagnosis of cardiovascular disease and no use of cardiological medications. Cardiac group (n = 30): Hospitalized adult patients with a confirmed diagnosis of cardiovascular disease. Data Collection: SCG signals were acquired using a smartphone with a triaxial accelerometer positioned on the lower third of the sternum. The Phyphox application was configured to record acceleration data on three axes (x, y, z) at a sampling rate of 100 Hz for 5 minutes per participant. Data Processing: Raw signals were processed using a custom pipeline developed in Google Colab, including bandpass filtering (2–39 Hz), beat detection, ensemble averaging, and extraction of temporal and amplitude parameters (heart rate, heart rate variability, respiratory rate, and fiducial points). Files Included: participants.xlsx — Spreadsheet containing demographic and clinical data for all participants (number, age, sex, group). data/ — Folder containing raw accelerometer signal files for each participant (1.csv, 2.csv, 3.csv, ..., 60.csv). Data Format: participants.xlsx: Microsoft Excel spreadsheet with four columns: number, age, sex, group. data/*.csv: CSV files containing raw accelerometer data with columns: time, x, y, z (acceleration in m/s²). Ethical Statement: All participants provided written informed consent. Data have been anonymized to protect participant privacy; no identifiable information is included. Related Platform: The signal processing algorithm developed in this study is available at: https://sismocardiography.streamlit.app/
Files
Steps to reproduce
1. Download the dataset files (participants.xlsx and data/ folder). 2. The participants.xlsx file contains demographic and clinical data for all 60 participants (number, age, sex, group). 3. The data/ folder contains raw accelerometer signals for each participant (1.csv to 60.csv). Each CSV file has columns: time (seconds), x, y, z (acceleration in m/s²). 4. To process the signals: a. Apply a bandpass filter (2–39 Hz) to remove artifacts and noise. b. Detect heartbeats using peak detection. c. Segment the signal into individual heartbeats. d. Calculate the ensemble average waveform for each participant. e. Extract temporal parameters: heart rate (HR), heart rate variability (SDNN, RMSSD), respiratory rate (RR), and fiducial points (MC, AO, AC, MO) with corresponding intervals (IVC, LVET, IVR, diastolic filling). 5. The processing pipeline is available at: https://sismocardiography.streamlit.app/ 6. For statistical analysis, use the Mann-Whitney test to compare groups (healthy vs. cardiac patients), with a significance level of 5% (p < 0.05).