Identification of adequate neurally adjusted ventilatory assist (NAVA) during systematic increases in the NAVA level

Abstract

Neurally adjusted ventilatory assist (NAVA) delivers airway pressure (P(aw)) in proportion to the electrical activity of the diaphragm (EAdi) using an adjustable proportionality constant (NAVA level, cm·H(2)O/μV). During systematic increases in the NAVA level, feedback-controlled down-regulation of the EAdi results in a characteristic two-phased response in P(aw) and tidal volume (Vt). The transition from the 1st to the 2nd response phase allows identification of adequate unloading of the respiratory muscles with NAVA (NAVA(AL)). We aimed to develop and validate a mathematical algorithm to identify NAVA(AL). P(aw), Vt, and EAdi were recorded while systematically increasing the NAVA level in 19 adult patients. In a multistep approach, inspiratory P(aw) peaks were first identified by dividing the EAdi into inspiratory portions using Gaussian mixture modeling. Two polynomials were then fitted onto the curves of both P(aw) peaks and Vt. The beginning of the P(aw) and Vt plateaus, and thus NAVA(AL), was identified at the minimum of squared polynomial derivative and polynomial fitting errors. A graphical user interface was developed in the Matlab computing environment. Median NAVA(AL) visually estimated by 18 independent physicians was 2.7 (range 0.4 to 5.8) cm·H(2)O/μV and identified by our model was 2.6 (range 0.6 to 5.0) cm·H(2)O/μV. NAVA(AL) identified by our model was below the range of visually estimated NAVA(AL) in two instances and was above in one instance. We conclude that our model identifies NAVA(AL) in most instances with acceptable accuracy for application in clinical routine and research.

Fig. 1.

Principles of neurally adjusted ventilatory assist (NAVA) . The diaphragm electrical activity (EAdi) derived from electrodes on a naso-gastric feeding tube is first amplified and processed. The EAdi signal is then multiplied by an adjustable gain factor (NAVA level) and used to control the pressure generator of a mechanical ventilator. Thus, NAVA delivers pressure to the airways () in direct synchrony and linear proportionality to the patient's neural inspiratory drive as reflected by the EAdi ( = EAdi() ⋅ (). Vt = tidal volume. = NAVA level that provides adequate unloading of respiratory muscles.

Fig. 2.

Example of a NAVA level titration session as used for estimating (a) visually or (b) with the proposed algorithm. refers to the adequate NAVA level early after the transition from the initial steep increase in and Vt(), referred to as 1st response, to the less steep increase or plateau in and Vt(), referred to as 2nd response –. Flow() is the air flow. In (a), the Vt() is estimated on a breath-by-breath basis. If there is false triggering of the ventilator (e.g., based on an EAdi artifact) a minimal Vt (normally a few milliliters) is delivered. Since there is no minimal threshold for Vt, the ventilator displays whatever Vt() is delivered in the graph. In (b), the Vt() is calculated as the integral of Flow() per inspiration as it is described in Section II-B (Step 4A).

Fig. 3.

Outline of the algorithm to identify based on the signals () for the NAVA level, for electrical activity of the diaphragm, and Vt for tidal volume that was derived from the inspiratory flow.

Fig. 4.

(a) Tracking of the NAVA level titration session in Patient 1 (Step 1). (b) Algorithm for modeling {( with lines (Step 1A).

Fig. 5.

() titration session tracking by 2 Gaussian components for Fig. 4. The component with small dispersion corresponds to Titration class (Step 1B).

Fig. 6.

The result of titration tracking procedure of Fig. 5. The lines that belong to are assigned to the Titration class (Step 1B).

Fig. 7.

Tracking of neural inspiration sessions using signal (Step 2C).

Fig. 8.

Clustering of frames to Neural Inspiration and Expiration classes (Step 2C).

Fig. 9.

The air flow signal, , is divided into inspirations and expirations by zero crossing indices (Step 4A).

Fig. 10.

A fuzzy logic factor used for exploiting bias to 0.25 of total duration of titration session (Step 4C).

Fig. 11.

Time index of plateau, , is found when is minimized, as described in Steps 3 and 4.

Fig. 12.

Comparison between independently estimated by one of the authors (L.B., a physician) and by 17 independent physicians based on visual inspection of the airway pressure () and tidal volume (Vt) response to systematic increases in the NAVA level (circles) and identified by the algorithm described in this paper (squares).

Fig. 13.

The graphic interface provides a synopsis of the signal processing steps described in Figs. 2, 5, 8, and 11, and allows for real time assessment of how the algorithm identifies . Ground truth denotes the visually estimated adequate NAVA level.

Fig. 14.

NAVA level titration session in patient 17. In this patient the algorithm identified the transition from a steep increase in peak airway pressure () to a less steep increase or plateau in (i.e., the adequate NAVA level, ) clearly below the range of as visually estimated by the clinicians. The discrepancy is most likely due to a short, transitory interruption of the increase during the initial steep increase, i.e., during the 1st response phase (asterisk). We assume that the physicians outperformed the current version of the algorithm in recognizing pattern irregularities.