Multi-view ultrasound for diaphragm monitoring and cough strength estimation

The current extubation process for mechanical ventilation is subjective and imprecise, as it lacks insight into respiratory muscle function. Ultrasound imaging has become an insightful technique to investigate diaphragmatic function, but common techniques have limitations in terms of repeatability and automation. We aim to develop a wearable ultrasound system to automatically quantify diaphragm movement from intercostal and subcostal views. We focus on diaphragm monitoring and cough strength estimation.

The proposed system consists of angle-free probes, a tracking algorithm, and a downstream algorithm. The probe allows for rapid setup and automatic diaphragm positioning. To provide real-time monitoring, we proposed a diaphragm tracking algorithm, which is automated and applicable for multiple ultrasound views. Sixty healthy participants (29 males and 31 females) were recruited to assess the system's accuracy and feasibility. In our validity study, we compared our measurements with those from a commercially available ultrasound scanner. Using sonomyographic techniques and machine learning regressors, we evaluated the system's potential for diaphragm monitoring and cough strength estimation in ICU applications.

Here, we show that our tracking algorithm exhibits strong resistance to motion interference, achieving a Jaccard index of 73.6% and a boundary accuracy of 84.8%. In the validity study, the wearable probes demonstrate minimal measurement error. Additionally, the system enables the implementation of monitoring and cough strength estimation, achieving an error of less than 50 L/min.

Using our advanced sonomyographic techniques, we observe distinct patterns of muscle activation and demonstrate the feasibility of analyzing the underlying mechanisms of complex respiratory movements. The results show the potential for these methods to serve as standard practice in the ICU for various applications during and after extubation.