Published October 6, 2026 | Version v1
Dataset Open

CACHET-CADB Mini: A Two-Recording Subset of the CACHET-CADB Database for Software Testing

Authors/Creators

  • 1. Technical University of Denmark
  • 2. Bispebjerg-Frederiksberg Hospital
  • 3. U. N. Mehta Institute of Cardiology and Research Centre
  • 4. ROR icon China Agricultural University

Description

This archive contains a small subset (2 recordings) of the CACHET-CADB database, assembled for automated unit and continuous-integration testing of the torch_ecg framework (https://github.com/DeepPSP/torch_ecg).

CACHET-CADB is a contextualized ambulatory electrocardiography arrhythmia dataset recorded with a wearable chest-patch ECG device (Kumar, Puthusserypady, Dominguez, Sharma & Bardram, Frontiers in Cardiovascular Medicine, 2022, DOI 10.3389/fcvm.2022.893090). Recordings come with cardiologist annotations (arrhythmia events) and contextual information. The full database (16.5 GB) is openly available at DTU Data (DOI 10.11583/DTU.14547264).

Contents (118 MB compressed / 353 MB uncompressed), in the original Unisens folder layout, unchanged from the full database:

  • CACHET-CADB-Mini/signal/P1/<participant>/<recording>/: ecg.bin (ECG, 1024 Hz), press.bin, marker.csv, heart-rate / HRV / accelerometer channels, and unisens.xml metadata.
  • CACHET-CADB-Mini/annotations/: the accompanying annotation and context files for the same recordings.

Why this subset exists: the full database is too large for CI test runs, and the hosting portal throttles scripted downloads, which made automated testing unreliable. This subset keeps the original file formats and folder structure so that parser code is exercised against real data.

Integrity:
MD5          d02b886831307bff37736377f0adc160
SHA-256   afd86755e3db33ac7db771d1fd8be2c881cab3300feebb63a8cfd678051fa86d

License: The underlying data belong to the CACHET group at DTU; no additional rights are granted by this redistribution. Refer to the original database license (DTU Data, DOI 10.11583/DTU.14547264) before use.

Citation: Please cite the original database paper (DOI 10.3389/fcvm.2022.893090) and this record.

Files

Files (124.2 MB)

Name Size Download all
md5:d02b886831307bff37736377f0adc160
124.2 MB Download