Subspace detection of the impulse response function from intra-partum cardiotocography

Area of Science:

• Biomedical Engineering
• Signal Processing
• Obstetrics

Background:

• Cardiotocography (CTG) records maternal uterine pressure (UP) and fetal heart rate (FHR) during labor.
• CTG data is often corrupted by noise and missing values, complicating analysis.
• Accurate modeling of CTG dynamics is crucial for monitoring fetal well-being.

Purpose of the Study:

• To model the input-output dynamics of CTG as an impulse response function (IRF).
• To evaluate the effectiveness of subspace methods for CTG data analysis, particularly with noisy and non-contiguous data.
• To compare the performance of subspace methods against linear regression in identifying pathological labor records.

Main Methods:

• Utilized subspace identification methods to model CTG data, incorporating noise suppression.
• Applied methods to both contiguous and non-contiguous CTG datasets.
• Employed linear regression as a benchmark for comparison.

Main Results:

• Subspace methods successfully modeled more pathological CTG records (30/57) compared to linear regression (26/57) using contiguous data.
• Allowing non-contiguous data further increased the number of modeled pathological records to 49 using subspace methods.
• The impulse response function (IRF) gain derived from subspace models showed statistically significant differences between normal and pathological records more frequently than linear regression (15/18 vs. 10/18 epochs).

Conclusions:

• Subspace methods offer a robust approach for modeling CTG data, effectively handling noise and missing data.
• This technique improves the identification of pathological labor cases compared to traditional linear regression.
• The enhanced discriminatory power of subspace-derived models aids in better assessment of fetal status during labor.