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

This is the authors' abstract. We don't add key points for this paper.

Mapping the ventilation and the regional blood pulsatility behavior of the lung is essential for treating patients with respiratory failure, especially those under mechanical ventilation. The separation of ventilation and blood pulsatility in Electrical Impedance Tomography images of the thorax is the focus of this paper, with potential applications in medical imaging and other tomographic imaging modalities. The method has two stages. In the first stage, the algorithm is trained to identify the dynamic models of ventilation and pulsatility cycles separately. The second stage uses the adjusted models to separate new incoming images in real-time. During the training stage, two average cycles are estimated - one for ventilation and the other for blood pulsatility. Coherent averages of a training set of images are computed using triggering signals to identify the cycles. These average cycles are then used to adjust evolution models for real-time processing in the second stage. The proposed method was evaluated with experimental data in swines under mechanical ventilation. The robustness of the method against substantial changes in ventilation modes was assessed. The results show that the method was successful in separating ventilation and pulsatility, with only a small residual mixture, and showed that the method is robust.