Tidal Volume Estimation Using Portable Ultrasound Imaging System

Tidal volume (inhaled air volume per breath) is an important index to indicate whether a person has been infected by respiratory diseases. In this paper, we evaluated a solution of tidal volume measurement based on an ultrasound system, which is more suitable for patient who cannot understand and cooperate with respiratory test. It also allows directive visualization of movement of respiratory muscles to assist doctor's diagnosis and treatment. The system estimates tidal volume only based on ultrasound video and pretrained model. It combines Viola-Jones and Chan-Vese algorithms to detect and segment diaphragm from ultrasound image sequences. Three statistical heuristics, such as principle direction, centroid of mass, and minimum-area-rectangle, are applied to automatically calculate the diaphragmatic excursion during one respiration cycle. Two experiments are conducted. The first experiment evaluated the performance of auto diaphragm detection. The accuracy is over 96%. The second experiment evaluated the performance of tidal volume estimation. The results verified a linear relationship between the diaphragmatic excursion and the tidal volume measured by a spirometer. When the whole diaphragm is shown in the ultrasound image, principle direction performed best and the average error between predicted tidal volume and ground truth from spirometer is about 0.1 L. Therefore, ultrasound device system is a feasible solution for tidal volume estimation.

Duke Scholars

Author Ming-Chun Huang DKU Faculty