PTB-XL+, a comprehensive electrocardiographic feature dataset

Abstract

Machine learning (ML) methods for the analysis of electrocardiography (ECG) data are gaining importance, substantially supported by the release of large public datasets. However, these current datasets miss important derived descriptors such as ECG features that have been devised in the past hundred years and still form the basis of most automatic ECG analysis algorithms and are critical for cardiologists' decision processes. ECG features are available from sophisticated commercial software but are not accessible to the general public. To alleviate this issue, we add ECG features from two leading commercial algorithms and an open-source implementation supplemented by a set of automatic diagnostic statements from a commercial ECG analysis software in preprocessed format. This allows the comparison of ML models trained on clinically versus automatically generated label sets. We provide an extensive technical validation of features and diagnostic statements for ML applications. We believe this release crucially enhances the usability of the PTB-XL dataset as a reference dataset for ML methods in the context of ECG data.

Fig. 1

Schematic overview of the components that constitute the PTB-XL + dataset.

Fig. 2

Lead- and segment-specific features as provided in the different feature sets. Color-coding corresponds to the fraction of samples for which values are present whereas black corresponds to values present for all samples. We report average statistics across leads X. The used acronyms are described in feature_description.csv.

Fig. 3

Global (sample-wise) ECG features as provided within the different feature sets. Color-coding as in Fig. 2.

Fig. 4

PTB-XL label distribution according to 12SL’s automatic diagnostic statements (showing the 40 most frequent statements out of overall 117 statements present in the whole dataset).

Fig. 5

Feature comparison based on (Pearson) correlation coefficients (left: 12SL vs. Uni-G, center: Uni-G vs. ECGDeli, right: 12SL vs. ECGDeli).

Fig. 6

Visual comparison of the label distribution for 12SL vs. original PTB-XL after mapping to SNOMED CT. On the x-axis we show the SNOMED CT labels ordered by ascending counts in the PTB-XL label set.