Beat-to-beat fetal heart rate analysis using portable medical device and wavelet transformation technique

Figures

• 

  Figure 1. Clinical simultaneous measurement of Baby Sound A pocket fetal doppler from Contec Medical [17] and CTG clinical device used from Bionet brand (FC-1400).

• 

  Figure 2. Statistical distribution of clinical data of pregnant women. The distributions of (a) Normal vs high risk pregnancies, (b) BMI of pregnant moms before pregnancy, (c) gestational pregnancy week, (d) gender of the fetus, are depicted.

• 

  Figure 3. A schematic representation of a single heartbeat cycle from an ultrasound probe. Temporal differences between heart sounds and ECG signals are shown.

• 

  Figure 4. Flowchart of the steps for our algorithm development for the fHR evaluation.

• 

  Figure 5. Distribution plot of RWEER for different 85 mother wavelets including Daubechies (db 1, db 2, …, db 45) , Symlets (sym 1, sym 2, …, sym 20), Coiflets (coif 1, coif 2, …, coif 5), and Biorthogonal (bior 1.1, bior 1.3, …, and bior 6.8), respectively. First row: The effect of wavelet decomposition in different levels on the abundance of maximum RWEER. It shows that level j = 4 works better than others. Second row: The investigation of RWEER for 215 sections with different mother wavelets on level j = 4, red dash-dot graph interprets the proposed algorithm's execution time for each mother wavelet, the black dash-dot graph represents the value of meanstd for shown box-plot. Outliers are drawn in red plus.

• 

  Figure 6. A sample of an original fHS signal before and after filtering: (a) Original signal, (b) Denoised signal using the proposed method, and (c) Extracted S-peaks. The red graph shows the obtained envelope and the red bullets represent all local maxima. Green bullets illustrate the position of selected S.

• 

  Figure 7. Single measurement data (signal of around 8 minutes) for the beat-to-beat cross-correlation analysis (Left plot), as well as the Bland-Altman plot for the obtained fHR by the proposed algorithm and the reference data (Right plot). Here SNR = −26.7 dB, and the p-value is <0.0001 for the simulated measurement data.

• 

  Figure 8. Statistical analysis over the performance of the algorithm is reported. The average p-value is <0.0008. Left: root mean square error (RMSE). Right: mean difference and standard deviation (STD) difference between calculated fHR and baseline for the simulated dataset are plotted.

• 

  Figure 9. The use of image processing techniques for validation of the proposed algorithm. a) Saved image from the clinical device. b, c) fHR graph extraction, d) Confidence interval definition, and e) Adding fHR results obtained from the proposed algorithm in red.

• 

  Figure 10. A comparison sample between the algorithm's final graph (in red) and the result of the clinical device (in black). The confidence interval is shown in green. Outliers are shown by red arrows.

• 

  Figure 11. The error obtained for each data. (a) The percentage of obtained error (orange graph) as well as the distribution of the median of calculated fHR for each data in bpm (purple bars). (b) The quantile distribution of the measured fHR values in bpm.

• 

  Figure 12. The results of the comparison of the current work with the works performed by others [35], [36].