Patient and database shift in lightweight ECG arrhythmia classification: a controlled evaluation of noise, quantization, and embedded inference
Medical Engineering & Physics · 2026
DOI: 10.1088/1873-4030/aeb4bc ↗Kurz gefasst
Fachartikel in der Zeitschrift Medical Engineering & Physics, 2026. Verfasst von Hyun Mok Park und 7 weiteren Autor:innen. Beteiligte Einrichtung: Digital Convergence Research Headquarters, KIEL Institute, Bucheon, Republic of Korea. Hauptthemen laut MEDLINE: Electrocardiography, Signal Processing, Computer-Assisted, Arrhythmias, Cardiac, Databases, Factual. Die Arbeit nennt 35 Literaturangaben. Gefördert von: Ministry of Trade, Industry and Resources.
Abstract (Originalsprache)
Efficient microcontroller execution does not ensure reliable electrocardiogram classification across patients and databases. We conducted a controlled benchmark of three convolutional architectures under a corrected subject-disjoint MIT-BIH split, external INCART testing, noise stress test database artifacts, 16-bit floating-point (FP16) and full-integer 8-bit (INT8) conversion, and classifier-only invocation on an ESP32 without external RAM. Five fixed seeds were used. On the MIT-BIH test cohort (DS2), INT8 four-class macro-F1 was 0.474-0.498 and supraventricular (S) recall was 0.150-0.180. Mean S-class area under the receiver operating characteristic curve (AUROC) was below 0.5 in all six architecture-regime groups, although all corresponding record-cluster intervals included 0.5. Within-record ranking could remain high despite poor pooled ranking. On INCART, S-class ranking was stronger, with all six patient-cluster AUROC intervals above 0.5, but the representative mixed-noise depthwise-separable convolutional neural network (DS-CNN) achieved 0.797 recall with 0.074 precision. Including record 201, which shares a subject with test record 202, increased DS2 S-recall point estimates by 2.49-14.68 percentage points. In the primary architecture panel, FP16 and INT8 met the engineering tolerances in all 120 comparisons. DS-CNN required 9.49 ms per classifier invocation on the ESP32 using preprocessed, annotation-centered inputs. Acquisition, QRS detection, preprocessing, current, and energy were not measured. Patient-dependentS-score behavior and external false-positive predictions limited this panel more than numerical precision or microcontroller resources.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Digital Convergence Research Headquarters, KIEL Institute, Bucheon, Republic of Korea.
ORCID-Profile: Hyun Mok Park, Woojin Cho
Förderung
- Ministry of Trade, Industry and Resources
Zitieren
APA
Park, H. M., Cho, W., Bang, S. O., Choi, H. S., Kwon, K. T., Roh, J. S., Shin, S. W., & Lim, J. M. (2026). Patient and database shift in lightweight ECG arrhythmia classification: a controlled evaluation of noise, quantization, and embedded inference. Medical Engineering & Physics. https://doi.org/10.1088/1873-4030/aeb4bc
Vancouver
Park HM, Cho W, Bang SO, Choi HS, Kwon KT, Roh JS, et al. Patient and database shift in lightweight ECG arrhythmia classification: a controlled evaluation of noise, quantization, and embedded inference. Medical Engineering & Physics. 2026. doi:10.1088/1873-4030/aeb4bc
BibTeX
@article{Park2026,
author = {Park, Hyun Mok and Cho, Woojin and Bang, Seok-Oh and Choi, Hyun-Seok and Kwon, Ki-Tae and Roh, Jin-Sung and Shin, Sang-Wuk and Lim, Jong-Min},
title = {Patient and database shift in lightweight ECG arrhythmia classification: a controlled evaluation of noise, quantization, and embedded inference},
journal = {Medical Engineering & Physics},
year = {2026},
doi = {10.1088/1873-4030/aeb4bc}
}
EUVIMED