A real-time method for ventilation and blood pulsatility separation in functional images of the chest

Figure 1.  Representative pulmonary air volume signal as a function of discrete time $ k $ and associated train of pulses ${p_\text{v}^\pm} $

Figure 2.  Eliminating damping and instabilities of the poles

Figure 3.  Representative sequence of images of the average pulsatility $ {\bar{\boldsymbol{P}}^\diamond_\text{T}}{}$ (left) and ventilation $ {\bar{\boldsymbol{V}}^\diamond_\text{T}}{}$ (right) obtained from EIT images of the first swine subject. Time increases in the clockwise direction. Not all images of the ventilation cycle are presented in this figure

Figure 4.  First eight principal components of pulsatility $ {\bar{\boldsymbol{P}}^\diamond_\text{T}}{}$ (left) and ventilation ${\bar{\boldsymbol{V}}^\diamond_\text{T}}{} $ (right) of the first swine subject. Each component is accompanied by the associated normalized singular value

Figure 5.  Discrete time sequence of the four principal components of the average pulsatility and ventilation cycles over one cycle of the first swine subject

Figure 6.  Comparison between the coefficients $ {\bar{\boldsymbol{\alpha}}^\text{v, p}_{{k}}}$ (dashed black lines) and the predicted values ${\hat{\boldsymbol{\alpha}}^\text{v, p}_{{k}}} $ (red) of the first subject. The gray plots display the relative error of the predictions

Figure 7.  Time history of the sum of the pixels of the images of the first subject. From top to bottom: Input, LMS total, Ventilation, Pulsatility, Offset, and Residual components, as defined in (27). The right column of plots presents zoomed plots of the highlighted regions on the left

Figure 8.  Time sequence of ventilation and pulsatility separation along one ventilation cycle of the first subject. From left to right: input images $ \boldsymbol{I_k} $, ventilation component $ \boldsymbol{I_k}^\text{v} $, and pulsatility component $ \boldsymbol{I_k}^\text{p} $ using two color maps, the first with the same color map and limits of the other images and the second using another colormap and auto-scale to increase contrast

Figure 9.  First four principal components of pulsatility ${\bar{\boldsymbol{P}}^\diamond_\text{T}}{} $ (left) and ventilation ${\bar{\boldsymbol{V}}^\diamond_\text{T}}{}$ (right) of the second swine subject. Each component is accompanied by the associated normalized singular value

Figure 10.  Time history of the sum of the pixels of the images of the second subject. From top to bottom, input, LMS total, ventilation, pulsatility, offset, and residual components as defined in (27). The left column displays the entire dataset and the next columns present zoomed plots of the colored highlighted regions on the left column. The red region is similar to the dynamics present in the training set, the green region has ventilatory frequency halved, the blue region has a ventilation pause, and the magenta has changes to the tidal volume and PEEP

Figure 11.  Region of interest of the heart (red), left, and right lungs (yellow) of the second subject, segmented by a specialist. Each area is composed of 6 pixels

Figure 12.  Time history of the sum of the pixels of the ROIs of the heart (top row), left lung (middle row), and right lung (bottom row.) in each of the colored highlighted segments of Figure 10 of the second subject. The black lines are the sum of the pixels of the input images in each ROI, while the colored plots are the separated ventilation and pulsatility. The pulsatility plot was presented separately in the grayed region for clarity