Table 2.

Columns provided in the metadata table ptbxl_database.csv.

Section	Variable	Data Type	Description
Identifiers	ecg_id	integer	unique ECG identifier
	patient_id	integer	unique patient identifier
	filename_lr	string	path to waveform data (100 Hz)
	filename_hr	string	path to waveform data (500 Hz)
General Metadata	age	integer	age at recording in years (see Fig. 3 left)
	sex	categorical	sex (male 0, female 1)
	height	integer	height in centimeters (see Fig. 3 right)
	weight	integer	weight in kilograms (see Fig. 3 middle)
	nurse	categorical	involved nurse (pseudonymized)
	site	categorical	recording site (pseudonymized)
	device	categorical	recording device
	recording_date	datetime	ECG recording date and time
ECG Statements	report	string	ECG report from diagnosing cardiologist
	scp_codes	dictionary	SCP ECG statements (see Tables 6, 7 and 8)
	heart_axis	categorical	heart’s electrical axis (see Table 10)
	infarction_stadium1	categorical	infarction stadium (see Table 11)
	infarction_stadium2	categorical	second infarction stadium (see Table 11)
	validated_by	categorical	validating cardiologist (pseudonymized)
	second_opinion	boolean	flag for second (deviating) opinion
	initial_autogenerated_report	boolean	initial autogenerated report by ECG device
	validated_by_human	boolean	validated by human
Signal Metadata	baseline_drift	string	baseline drift or jump present
	static_noise	string	electric hum/static noise present
	burst_noise	string	burst noise
	electrodes_problems	string	electrodes problems
	extra_beats	string	extra beats
	pacemaker	string	pacemaker
Cross-validation Folds	strat_fold	integer	suggested stratified folds

Each ECG is identified by a unique ID (ecg_id) and comes with a number of ECG statements (scp_codes) that can be used to train a multi-label classifier that can be evaluated based on the proposed fold assignments (strat_fold).