https://github.com/lezaf/cardiotocography-signals-analysis

Biomedical analysis of Heart Rate Variability (HRV) of fetus relationship with various umbilical artery blood measurements. Correlation analysis and machine learning techniques are applied to detect abnormalities.
https://github.com/lezaf/cardiotocography-signals-analysis

biomedical-signal-processing machine-learning scipy signal-processing sklearn

Last synced: about 1 month ago
JSON representation

Biomedical analysis of Heart Rate Variability (HRV) of fetus relationship with various umbilical artery blood measurements. Correlation analysis and machine learning techniques are applied to detect abnormalities.

• Host: GitHub
• URL: https://github.com/lezaf/cardiotocography-signals-analysis
• Owner: lezaf
• License: mit
• Created: 2024-06-25T16:26:46.000Z (5 months ago)
• Default Branch: master
• Last Pushed: 2024-06-25T18:38:19.000Z (5 months ago)
• Last Synced: 2024-10-02T06:03:09.929Z (about 1 month ago)
• Topics: biomedical-signal-processing, machine-learning, scipy, signal-processing, sklearn
• Language: Python
• Homepage:
• Size: 277 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# cardiotocography-signals-analysis

We performed ML analysis on intrapartum cardiotocography (CTG) signals. Here is a brief summary of the [data](#data), the [preprocessing](#preprocessing) we applied on signals, the [analysis](#process) methods and some of the [results](#Snapshots).

## Data

We used intrapartum cardiotocography (CTG) signals [database](https://www.physionet.org/content/ctu-uhb-ctgdb/1.0.0/) from PhysioNet. Each CTG contains a FHR time series and a Uterine Contraction (UC) signal. The database also includes maternal, delivery, and fetal clinical details.

## Preprocessing

To handle noise in CTG signals we performed following steps:

- Step 1: Leading/trailing zeroes elimination

- Step 2: Intermediate zero parts

  - We set a time threshold of *7 seconds* for acceptance of intermediate zero parts. Above this threshold the intermediate parts were eliminated since their interpolation would give an unrealistic behavior of the signal.

- Step 3: Remaining zero values interpolation

- Step 4: Extremely low/high HR values interpolation

The preprocessing effect on CTG signals is depicted below:



## Process

We used Fetal Heart Rate (FHR) time series to calculate Heart Rate Variability (HRV) of fetus. Performing correlation analysis between HRV and the available features, we concluded to a subset of features with underlying relationship:

- pH

- BDecf

- BE

For the case of HRV/pH relationship we further experiment with ML algorithms, including *k-means* and *DBSCAN* to decide normal/abnormal pH range.

## Snapshots

For the case of HRV/pH in *vaginal* delivery, we concluded in *normal* pH values: $7.17 \leq pH \leq 7.32$.



## Citations

- [Václav Chudáček, Jiří Spilka, Miroslav Burša, Petr Janků, Lukáš Hruban, Michal Huptych, Lenka Lhotská. Open access intrapartum CTG database. BMC Pregnancy and Childbirth 2014 14:16](https://bmcpregnancychildbirth.biomedcentral.com/counter/pdf/10.1186/1471-2393-14-16.pdf)

- Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., Ivanov, P. C., Mark, R., ... & Stanley, H. E. (2000). PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation [Online]. 101 (23), pp. e215–e220