M.Sc. Silvia Becker

M.Sc. Ekaterina Klepikova

Motivation
Atrial cardiomyopathy (AtCM) is associated with slow-conducting low-voltage areas and therefore prolonged total atrial conduction time. Jadidi et al. demonstrated that total atrial conduction time correlates with amplified P-wave (APW) duration. Furthermore, APW morphology has been demonstrated to assist in staging AtCM. These findings suggest that APW duration and morphology measured in a 12-lead-ECG enables the noninvasive diagnosis of AtCM. Despite its diagnostic value, manual annotation of APW morphology is time-consuming, subjective, and not scalable, rendering it impractical for large-scale screening or retrospective cohort studies. To overcome these limitations, we aim to develop an automated, AI-based algorithm that can annotate APW morphology reliably and reproducibly.

Student Project
The research question we aim to answer is: Can an AI-based algorithm accurately detect and classify amplified P-wave morphology (biphasic, late P, . . . ) from ECG signals to support the diagnosis and risk stratification of atrial cardiomyopathy? Within this project, a CNN-based model will be developed for a multi-label classification task, targeting the following P-wave morphologies:
• Biphasic P-wave in lead II
• Biphasic P-wave in lead III
• Biphasic P-wave in lead aVF
• Late P

The project will systematically investigate different data preprocessing and augmentation strategies, as well as three input formats:
• 12-lead single-beat ECG template
• Single-beat ECG template with 3 leads only
• Full 10 s resting ECG

ECG data will be sourced from open datasets and the University Heart Center Freiburg - Bad Krozingen.