Epycon: A Single-Platform Python Package for Parsing and Converting Raw Electrophysiology Data into Open Formats

Jakub Hejc1, Richard Redina2, Jana Kolarova3, Zdenek Starek4
1International Clinical Research Center, St. Anne's University Hospital, Brno, Czech Republic; Department of Pediatric, Children's Hospital, The University Hospital Brno, Brno, Czech Republic, 2Brno University of Technology; International Clinic Research Centre, St. Anna's University Hospital, Brno, 3Brno University of Technology, 4International Clinical Research Center, St. Anne's University Hospital, Brno, Czech Republic; 1st Department of Internal Medicine, Cardio-Angiology, Faculty of Medicine, Masaryk University, Brno, Czech Republic

Abstract

Background: Integrated acquisition systems are used to support interventional cardiac electrophysiology procedures. However, internal tools for exporting data are often limited in their settings (channels settings, window of interest). Full access to raw data can be challenging due to undocumented structure of the underlying binary formats. Therefore, we aimed to develop an open-source platform for parsing and processing of electrophysiology recordings.

Methods: The structure of the proprietary binary files used by the WorkMate (Abbott) acquisition system was decoded through visual reverse engineering and inter-file pattern matching. The process included the identification of byte blocks containing the unfiltered data stream, acquisition settings, visualization parameters, and clinical annotations. The package was implemented in Python 3.8 allowing parsing the data and converting it into established open formats Comma-separated text format (CSV) and Hierarchical Data Format 5 (HDF5). The structure of HDF5 was made compatible with the freely accessible signal processing platform SignalPlant. Epycon supports memory mapping and lazy reading of acquired signals. It contains tools for export automation, procedure-specific lead mounting, or direct import into NumPy arrays.

Results: We tested the single-thread read-write performance of the parser using a dataset of 12 consecutive studies (mean size 4421.1±2635.3 MB) acquired during complex animal procedures. Each study was performed with identical acquisition and channel settings (2000 Hz, 78 uV/LSB, 51 unmounted channels). The mean conversion time per study was 13.9±8.0 s and 178.4±106.6 s for HDF5 and CSV. The relative change in file size after conversion with preserved 32-bit precision was +2.1% (HDF5) and +97.0% (CSV).

Conclusions: The package provides core, memory-efficient functionality for handling the data acquired by the WorkMate system. The supported software versions include WorkMate 4.1–4.3. The open-format features allow researchers to work with full electrophysiology recordings using common computing platforms (Python, Matlab) or third-party visualization and processing software SignalPlant.