Classification of Cervical Dilation Stages through Combined EHG and Clinical Features: A Genetic Algorithm-Optimized Approach
Classification of Cervical Dilation Stages through Combined EHG and Clinical Features: A Genetic Algorithm-Optimized Approach
No Thumbnail Available
Date
2025
Authors
Otniel Portillo-Rodriguez, 0000-0002-2198-477X
OSCAR OSVALDO SANDOVAL_GONZALEZ, 0000-0002-0309-5231
Eric Alonso Abarca-Castro, 0000-0002-2029-3790
Juan C. Echeverría Arjonilla, 0009-0004-5625-1949
José Javier Reyes-Lagos, 0000-0001-5361-5007
Hugo Mendieta Zeron, 0000-0003-3492-8950
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad Autónoma Metropolitana. Unidad Lerma
Abstract
Description
This open dataset contains feature vectors derived from abdominal electrohysterogram (EHG) recordings collected between 2017 and 2019 from seventy-three healthy parturient women, aged eighteen to thirty-four years and bearing pregnancies of 30.5 to 41.2 weeks’ gestation. Recordings were made at the Maternal and Childhood Research Center (CIMIGen) in Mexico City and at the Maternal-Perinatal Hospital “Mónica Pretelini-Sáenz” in Toluca, Mexico, using the Monica AN24 portable maternal–fetal monitor. Raw signals were acquired at 900 Hz, automatically resampled to 20 Hz and band-limited to 0.2 – 1 Hz before export.
To preserve temporal information while enlarging the training set, every ten-minute trace was sliced into overlapping 120-second windows that slide forward in 60-second steps. This sliding-window strategy yielded 648 partially repeated samples. Each window was then split into three frequency bands: the complete 0.2 – 1 Hz band (B1), a high sub-band that emphasises the fast-wave-high component 0.34 – 1 Hz (B2) and a lower sub-band that emphasises the fast-wave-low component 0.2 – 0.34 Hz (B3).
Seven signal descriptors were calculated in every band, giving twenty-one EHG features per window. Acronyms follow the pattern “metric plus band index”. • RMS1, RMS2, RMS3: Root Mean Square amplitude in B1, B2 and B3. • AUC1, AUC2, AUC3: Area Under the rectified Curve in B1, B2 and B3. • SampEn1, SampEn2, SampEn3: Sample Entropy, a measure of complexity, for each band. • MNF1, MNF2, MNF3: Mean (spectral) Frequency in each band. • MDF1, MDF2, MDF3: Median Frequency that splits spectral power in half for each band. • ZCR1, ZCR2, ZCR3: Zero-Crossing Rate, the number of sign changes per second, for each band. • BubbEn1, BubbEn2, BubbEn3: Bubble Entropy, another complexity metric, for each band.
Four additional columns provide clinical context and are independent of any band-pass filtering. • MA: Maternal Age in years. • GA: Gestational Age in weeks. • LC: Low-amplitude uterine Contractions counted within the 120-second window. • HC: High-amplitude uterine Contractions counted in the same window.
Two label fields complete every record. • Dilation: cervical dilation measured clinically in centimetres. • Dilation Class: labour stage derived from Dilation, coded as Low for 1–4 cm, Moderate for 5–6 cm and Advanced for 7–10 cm.
Each row in the CSV file therefore contains twenty-seven columns: twenty-one band-specific EHG features, four clinical descriptors and the two dilation labels. The file, named ehg_cervical_dilation_features.csv, is encoded in UTF-8 and is distributed through the Universidad Autónoma Metropolitana institutional repository under a Creative Commons Attribution 4.0 International licence.
The raw EHG recordings can be downloaded from: https://xogi.ler.uam.mx/items/c142016c-87bd-4f71-8315-459e41e42418
Keywords
MEDICINA Y CIENCIAS DE LA SALUD