Open-source, vendor-independent, automated multi-beat tissue Doppler echocardiography analysis

Figures

• 

  Fig. 1. Acquisition procedure. Schematic showing the process of acquisition using a video capture device connected to any ultrasound machine with a VGA output port (top panel). Image area and scale are extracted from acquired snapshots (middle panel) and reconstructed to obtain an extended tissue Doppler spectrum (bottom panel)

• 

  Fig. 2. Processing steps for automated tracing and extracting peak velocities. The reconstructed tissue Doppler trace (top panel) is thresholded to separate foreground pixels (trace) from background pixels (second panel). Each column of this image (yellow bar) is analyzed to find the location of the highest filtered intensity (third panel schematic—left) which is defined as the middle of the envelope (blue circle), and largest drop in pixel intensity which is defined as the outer edge of the envelope (red circle). The outer edge (red) and middle (blue) traces are shown overlaid on the Doppler trace after analyzing each vertical column (third panel—right). The bottom panel shows the extended Doppler spectrum with automated outer (red) and middle (blue) traces with peak systolic (s′) and diastolic (e′ and a′) velocities (red and blue squares)

• 

  Fig. 3. Average velocity estimates of experts versus automated algorithm. Average septal s′, e′ and a′ velocity estimates and standard deviations at the middle of the Doppler envelope across 90 s using the experts’ consensus (red) and the automated algorithm (black) for each patient. The markers represent average velocity across 3 × 30 s recordings (94 ± 32 beats per patient) and error bars represent standard deviation for those beats. The patients have been placed in ascending order of velocity

• 

  Fig. 4. Bland–Altman plots for patient-by-patient analysis at the septal annulus. Bland–Altman plots comparing manual estimates from three experts and the automated algorithm’s estimates for septal peak s′, e′ and a′ velocities on a patient-by-patient basis

• 

  Fig. 5. Bland–Altman plots for patient-by-patient analysis at the lateral annulus. Bland–Altman plots comparing manual estimates from three experts and the automated algorithm’s estimates for lateral peak s′, e′ and a′ velocities on a patient-by-patient basis

• 

  Fig. 6. Modified Bland–Altman plots comparing manual estimates from three experts and the automated algorithm’s estimates for septal peak s′, e′ and a′ velocities on a beat-by-beat basis. An alpha-transparency of 0.1 has been applied to the plotted markers

• 

  Fig. 7. Bland–Altman plots for beat-by-beat analysis at lateral annulus. Modified Bland–Altman plots comparing manual estimates from three experts and the automated algorithm’s estimates for lateral peak s′, e′ and a′ velocities on a beat-by-beat basis. An alpha-transparency of 0.1 has been applied to the plotted markers

• 

  Fig. 8. Outer edge versus middle of envelope. Modified Bland–Altman plots comparing two different measurement conventions for septal s′ measurements: outer edge (top panels) and middle (bottom panels) of the Doppler envelope. An alpha-transparency of 0.1 has been applied to the plotted markers