Detection of T Wave Peak for Serial Comparisons of JTp Interval

Figures

• 

  FIGURE 1. Examples of the JTp and JT50 measurements. The rows (A–C) of to three different ECG recordings made in a male (32 years), female (22 years), and female (42 years), respectively. The underlying heart rates were 68, 95, and 96 beats per minute, respectively. In the left panels, representative P-QRS-T complexes are shown for all 12 leads superimposed on the same isoelectric axis, the middle panels show derived orthogonal XYZ leads also on the same isoelectric axis, and the right panels show the vector magnitude of the orthogonal XYZ leads. The horizontal (time) scale of all the panels is the same. The red, yellow, and green lines show the positions of the QRS onset, QRS offset, and T offset points. In the left panels, the violet line shows the T wave peak while the red-filled area (between the yellow and brown lines) is the first middle of the area under the T wave representation (i.e., the brown line divides the area under the T wave between the yellow and green lines into two equal halves). Measurements of the JTp and JT50 intervals are shown in the left panels. Note that JTp > JT50 in panel (A) while JTp < JT50 in panels (B,C) albeit the difference between JTp and JT50 is minimal…

• 

  FIGURE 2. Sample density distributions of the differences between JTp intervals measured based on different one-dimensional matrices in all ECG samples. The top, middle, and bottom panels show the differences between JTp intervals measured based on all other matrices and the measurements based on the XYZ, RMS, and SVD(QT) matrix, respectively. The graphs were constructed using 1 ms bins. Note that in some cases, the lines were superimposed and that some of the graphs might be hidden bellow others – e.g., within the precision of the graphics, the results of RMS(8) were practically identical to those of SVD(JT) which is shown on the top.

• 

  FIGURE 3. Sample density distributions of the differences between JT50 intervals measured based on different one-dimensional matrices in all ECG samples. The top, middle, and bottom panels show the differences between JT50 intervals measured based on all other matrices and the measurement based on the XYZ, RMS, and SVD(QT) matrix, respectively. The graphs were constructed using 1 ms bins. Note that in some cases, the lines were superimposed and that some of the graphs might be hidden bellow others – e.g., within the precision of the graphics, the results of RMS(8) were practically identical to those of SVD(JT) which is shown on the top.

• 

  FIGURE 4. Sample density distributions of the differences between JTp and JT50 intervals measured based on different one-dimensional matrices in all ECG samples. The graphs were constructed using 1 ms bins. Note that in some cases, the lines were superimposed and that some of the graphs might be hidden bellow others – e.g., within the precision of the graphics, the results of RMS(8) were practically identical to those of SVD(JT) which is shown on the top.

• 

  FIGURE 5. Bland-Altman plots of the individual differences between JTp and JT50 intervals based on XYZ matrix (top panel), RMS matrix (middle panel), and SVD(QT) matrix (bottom panel). In each panel, the bold red line shows the mean difference, the dashed red lines show the mean ± standard deviation. In each panel, the large blue marks show the means of the JTp – JT50 differences in 10-millisecond bins of the JTp and JT50 averages (from 90–100 ms to 290–300 ms); the blue error bars are the standard deviations of the JTp – JT50 differences in these bins.

• 

  FIGURE 6. Bland-Altman plots of the differences between curvilinear JTp/RR regression residuals in individual study subjects. The top, middle, and bottom panels show the differences for JTp interval based on XYZ and RMS matrices, XYZ and Quasi-orthogonal matrices, and XYZ and SVD(QT) matrices, respectively. In each panel, the red circles and blue squares show female and male subjects, the bold violet line shows the mean difference and the dashed violet lines show the mean ± standard deviation of the differences.

• 

  FIGURE 7. Bland-Altman plots of the differences between curvilinear JT50/RR regression residuals in individual study subjects. The top, middle, and bottom panels show the differences for JT50 interval based on XYZ and RMS matrices, XYZ and Quasi-orthogonal matrices, and XYZ and SVD(QT) matrices, respectively. In each panel, the red circles and blue squares show female and male subjects, the bold violet line shows the mean difference and the dashed violet lines show the mean ± standard deviation of the differences. Compare with Figure 6 and note the difference in the spread of the data.

• 

  FIGURE 8. Statistical summaries of the curvilinear JTp/RR (top panel) and JT50/RR (bottom panel) regression residuals for the interval measurements based on different one-dimensional ECG matrices. For comparison, the residuals of the QT/RR and JT/RR regressions are also shown in both panels. The red, blue, and violet bars show the mean values in female, males, and the complete study population, respectively. The errors bars show the corresponding standard deviations.

• 

  FIGURE 9. Cumulative sample distributions of the within-subject curvilinear residuals the JTp/RR (top panel) and JT50/RR (bottom panel) regressions superimposed for different ECG matrices. Note that while the red line of the XYZ matrix is on the left side of the other distributions, the differences are not particularly large (apart from the precordial matrix). Note that in some cases, the lines were superimposed and that some of the graphs might be hidden bellow others – e.g., within the precision of the graphics, the results of RMS(8) were practically identical to those of SVD(JT) which is shown on the top.

• 

  FIGURE 10. Distribution of the one-dimensional ECG matrices that led to the intra-subject lowest residuals for the JTp/RR (top panel) and JT50/RR (bottom panel) regressions. The pie charts show the proportions of the subjects in whom the given one-dimensional matrix led to the lowest regression residual.