PhysioNet/CinC Challenge 2016: Training Sets (2025)

Heart sound recordings were sourced from several contributors aroundthe world, collected at either a clinical or nonclinical environment,from both healthy subjects and pathological patients. The Challengetraining set consists of five databases (A through E) containing atotal of 3,126 heart sound recordings, lasting from 5 seconds to justover 120 seconds. You can browse these files or download the entire training set as a zip archive (169 MB).

In each of the databases, each record begins with the same letter followed by a sequential, but random number. Files from the same patient are unlikely to be numerically adjacent. The training and test sets have each beendivided so that they are two sets of mutually exclusive populations(i.e., no recordings from the same subject/patient were are in bothtraining and test sets). Moreover, there are two data sets that havebeen placed exclusively in either the training or test databases (toensure there are ‘novel’ recording types and to reduce overfitting onthe recording methods). Both the training set and the test setmay be enriched after the close of the unofficialphase. The test set is unavailable to the public and will remainprivate for the purpose of scoring.

Participants may note the existence of a validation dataset in the data folder.This data is a copy of 300 records from the training set, and will beused to validate entries before their evaluation on the test set. Moredetail will be provided in the scoring section below.

The heart sound recordings were collected from different locations onthe body. The typical four locations are aortic area, pulmonic area,tricuspid area and mitral area, but could be one of nine differentlocations. In both training and test sets, heart sound recordings weredivided into two types: normal and abnormal heart soundrecordings. The normal recordings were from healthy subjects and theabnormal ones were from patients with a confirmed cardiacdiagnosis. The patients suffer from a variety of illnesses (which wedo not provide on a case-by-case basis), but typically they are heartvalve defects and coronary artery disease patients. Heart valvedefects include mitral valve prolapse, mitral regurgitation, aorticstenosis and valvular surgery. All the recordings from the patientswere generally labeled as abnormal. We do not provide more specificclassification for these abnormal recordings. Please note that bothtraining and test sets are unbalanced, i.e., the number of normalrecordings does not equal that of abnormal recordings. You will haveto consider this when you train and test your algorithms.

Both healthy subjects and pathological patients include both childrenand adults. Each subject/patient may have contributed between one andsix heart sound recordings. The recordings last from several secondsto up to more than one hundred seconds. All recordings have beenresampled to 2,000 Hz and have been provided as .wav format. Eachrecording contains only one PCG lead.

Please note that due to the uncontrolled environment of therecordings, many recordings are corrupted by various noise sources,such as talking, stethoscope motion, breathing and intestinalsounds. Some recordings were difficult or even impossible to classifyas normal or abnormal.

 Name Last modified Size Description Parent Directory -  training-a/ 2016-09-03 00:02 -  training-b/ 2017-04-25 16:13 -  training-c/ 2018-11-09 13:29 -  training-d/ 2017-04-25 16:13 -  training-e/ 2017-04-25 16:13 -  training-f/ 2016-09-02 16:43 -  training.zip 2016-05-18 17:17 181M  validation.zip 2016-03-18 15:33 18M  validation/ 2016-03-20 00:01 -