Modelling ventilation with spontaneous breaths: Improving accuracy with shape functions and slice method.

BACKGROUND AND

OBJECTIVE

Accurate detection of spontaneous breathings (SBs) and respiratory asynchronies during mechanical ventilation (MV) is essential for optimizing patient care and preventing lung injuries. Conventional models often fail to capture these events with sufficient accuracy. To address this gap, this study introduces new equations incorporating custom shape functions and the Slice method, aiming to deliver a more robust, "bedside" model with potential applications in real-time asynchrony detection.

METHODS

Three new equations were developed to incorporate shape functions accounting for pressure- and volume-dependent changes in elastance, and a fourth model combined these shape functions with the Slice method. Retrospective data from 8 ICU patients (each providing 6 mins of ventilatory data) were split into two datasets of 4 patients each: one for model development and refinement, and the other for testing performance in reproducing ventilatory waveforms. Model accuracy was assessed using the coefficient of determination (R2) and Mean Residual Error (MRE). This evaluation focused on how effectively each model captured actual patient breathing mechanics, particularly in the presence of SBs or respiratory asynchronies.

RESULTS

The proposed models, especially the one combining shape functions with the Slice method-Recruitment Distention Elastance Analysis + Slice (RDEA + Slice)-exhibited a strong correlation with patient data, evidenced by high R2 values. While conventional models achieved R2 coefficients between 0.25 and 0.87, the new models improved these to 0.90-0.97. The RDEA + Slice model attained significantly lower MRE values (0.012-0.032), underscoring its superior accuracy in capturing dynamic changes. Furthermore, a unique identifiability analysis confirmed that the model parameters can be reliably estimated, supporting its potential for clinical application.

CONCLUSIONS

The new bedside models, especially RDEA + Slice, demonstrate promise in enhancing mechanical ventilation management. By accurately capturing ventilatory mechanics in presence of SBs, they hold potential to refine ventilator settings, reduce lung injury risks, and integrate with real-time diagnostic tools for detecting patient-ventilator asynchronies-ultimately supporting more personalized and effective ICU care.