$P$-waves are a form of first wave development in ECG signals that have substantial atrial medical information. Analysing P-waves with manual inspection is difficult because P-waves are small, vary and have a noisy appearance. Automatic classification of P-waves to detect atrial abnormalities is necessary to assist clinicians with faster process. This paper presents a P-wave morphological analysis using a random forest classification from 134 patients. The algorithm defines the data into five classes, namely, Normal, Left Atrial enlargement (LAE), Right Atrial Enlargement (RAE), Biatrial Enlargement (BE) and Atrial Fibrillation (AFib). This study uses ECG Lead II data from 12 standard medical leads. Signal processing and denoising are applied by using two filters, a derivative and Butterworth filter. Feature extraction is explored by using a new local distance transform, which is more efficient than other similar methods. The features used are P-wave morphological attributes such as duration, amplitude, number of appearances, standard deviation, and symmetry. The overall accuracy of our approach was 94.77%, the specificity (SP) was 98%, while the sensitivity (Se) at 10-fold validating the training set was 930%. This result comparable to other best performing algorithms and might be considered a second opinion for cardiologists.

Original languageEnglish
Title of host publicationProceedings - 2020 6th International Conference on Science and Technology, ICST 2020
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781728194721
Publication statusPublished - 2020
Event6th International Conference on Science and Technology, ICST 2020 - Yogyakarta, Indonesia
Duration: 7 Sept 20208 Sept 2020

Publication series

NameProceedings - 2020 6th International Conference on Science and Technology, ICST 2020


Conference6th International Conference on Science and Technology, ICST 2020


  • Atrial Fibrillation (AFib)
  • Atrium Enlargement
  • P-wave detection
  • P-wave morphology
  • electrocardiogram (ECG)
  • local distance transform
  • random forest classification


Dive into the research topics of 'Classification of P-wave Morphology Using New Local Distance Transform and Random Forests'. Together they form a unique fingerprint.

Cite this