Summary
There are various approaches to categorize fetal heart monitoring using classification models. One common approach is to extract features from fetal heart rate signals acquired through cardiotocography. These features can encompass time domain features (e.g. fetal heart rate baseline, acceleration, deceleration), frequency domain features (e.g. low/high frequency power), and nonlinear features (e.g. approximate entropy). A second approach is to input fetal magnetocardiography recordings. Independent component analysis can then be used to separate the recordings into fetal and non-fetal components. Algorithms identify the fetal components by analyzing frequency and temporal characteristics.A third approach is to combine quantitative parameters and signal images as inputs to a hybrid neural network to classify healthy and pathological fetuses. Another approach is to select features using methods like linear discriminant analysis, recursive feature elimination, forward/backward elimination, and lasso regression. The extracted features can then be used to train a classifier like support vector machines, neural networks, decision trees, etc. These approaches have been shown to achieve good classification performance, with some studies reporting over 80% accuracy.
Background: This study used the CTU-UHB database to analyze fetal heart rate data. Extracted time domain features: fetal heart rate baseline, acceleration, deceleration, long/short variations.Frequency domain features: low/high frequency power.Nonlinear features: approximate entropy. 21 features were extracted and classified using SVM with accuracy of 90.9% on test data.
Published By:
Y. Que - undefined
2020
Cited By:
0
An automatic method to retrieve fetal cardiac signal from fetal magnetocardiography (fMCG) recordings was developed. The method uses independent component analysis (ICA) to separate 66 multichannel fMCG datasets into fetal and non-fetal components. Algorithms then identify the fetal components by analyzing frequency and temporal characteristics. The automatic method was validated against manual classification by an expert. ICA was run with different input cluster sizes to simulate various MCG systems. The automatic method had high detection rates, low false positives, and reconstructed fetal signals with QRS amplitudes, standard deviations and signal-to-noise ratios comparable to manual classification. QRS differences between automatic and manual methods were small. The method was most robust with large input clusters. The automatic method could increase the diagnostic value of fMCG.
Published By:
D. Mantini - Physics in Medicine and Biology
2005
Cited By:
12
AI system achieved 80.1% accuracy in classifying fetal health from heart rate and contraction data.
Published By:
Edoardo Spairani - Frontiers in Bioengineering and Biotechnology
2022
Cited By:
5
The multivariate method based on fetal heart rate analysis can identify late intrauterine growth in two sentences.
Published By:
N. Pini - Frontiers in Artificial Intelligence
2021
Cited By:
14
EOG detects driver drowsiness by analyzing eye movements.Aggregating multiple eye features improves drowsiness prediction.
Published By:
P. Ebrahim - undefined
2016
Cited By:
14