(B-46) Oxygen Saturation effect on Respiratory Regulation

Introduction:: Physiological system identification is a crucial tool in understanding the dynamics that govern complex physiological functions. For about two decades, new techniques have been developed in our lab to advance this field; notably, the Laguerre expansion technique (Marmarelis, 2004). The technique uses the set of Laguerre orthogonal basis functions for efficient representation (small number of coefficients) that improves estimation accuracy for short and noisy data-records. In this abstract, we explore kernel representations of a new respiratory model for cognitively normal and impaired subjects in order to advance our understanding of how changes in blood gases influence the dynamics of respiratory regulation. This scientific/clinical question attained new importance during the recent Covid pandemic.


Materials and Methods:: Contemporaneous end-tidal CO2(EtCO2), respiration rate (RR), and oxygen saturation (SpO2) are measured continuously over 10 minutes for each subject (Session 1) and 10 minutes after a paced breathing session (Session 3) . The raw data is then high-passed at 0.005 Hz before the EtCO2, RR, and SpO2 values are extracted and, subsequently, resampled at 2 Hz with a cubic spline interpolation method. The Laguerre expansion technique (LET) is then applied to estimate the kernels of the SpO2-to-RR and CO2-to-RR dynamic relationships (RR is viewed as the output, and SpO2 and EtCo2 as the inputs of a linear time-invariant model). Principal dynamic modes (PDM) for each input are then computed by taking the singular value decomposition of the kernels for all the controls. The model is then computed by taking the PDM as the basis functions and recording the gain for each PDM.


Results, Conclusions, and Discussions:: The estimated set of kernels were obtained from 13 cognitively normal (NC) subjects and 20 cognitively impaired (MCI) patients. The kernels for each input were averaged over all subjects. It is notable to see that in Figure 1 the O2 average kernel has mostly negative values, indicating reduction of RR with increasing SpO2 (as anticipated for homeostatic operation). The kernel of the MCI patients, however, indicates an increase in RR for the early lags of the kernel (< 3 sec) and, more importantly, for later lags ( >8 seconds). This suggests a serious risk of dysregulation of blood SpO2 in MCI patients, preventing homeostasis (i.e. restoration of normal blood SpO2 levels). The CO2 kernels on the other hand (Figure 2), do not have any notable delineating features. Session 3 has nearly identical kernels for controls and MCI. The effect of these results can be noted in the gains of the PDMs.

The results showed distinct differences in the respiratory response to SpO2 changes of the two groups, with the cognitively impaired subjects exhibiting a serious risk of respiratory dysregulation compared to the normal subjects. This can also be seen with the difference in the gain of the first PDM for SpO2. The peak of PDM 1 for SpO2 at around 0.05 Hz is indicative of cholinergic signaling. We hypothesize that this is due to impaired signaling of blood oxygen in the respiratory chemoreflex that would normally induce a lower respiration rate with the increase of oxygen. Through the PDM analysis we can also see that the observed impairment in the patients’ O2 kernel can be due to a dysfunction of the vagal component of respiratory regulation. These findings have important implications for understanding the underlying physiological mechanisms of respiratory control in both normal and pathological conditions, including lingering questions in long Covid patients.


Acknowledgements (Optional): :

References (Optional): :