Evaluating the Nonlinear Dynamics of Cerebrovascular Reactivity with Transcranial Doppler Ultrasound

1857 

Abstract Submission 

Genevieve Hayes¹, Sierra Sparks¹, Joana Pinto¹, Daniel Bulte¹

¹University of Oxford, Oxford, Oxfordshire

Genevieve Hayes  
University of Oxford
Oxford, Oxfordshire

Sierra Sparks  
University of Oxford
Oxford, Oxfordshire

Joana Pinto, PhD  
University of Oxford
Oxford, Oxfordshire

Daniel Bulte, Associate Professor  
University of Oxford
Oxford, Oxfordshire

Cerebrovascular reactivity (CVR) describes the dilation and constriction of cerebral blood vessels in response to vasoactive stimuli such as altered PaCO2 [1,2]. CVR can be reproducibly assessed using transcranial Doppler ultrasound (TCD) of the middle cerebral artery (MCA) [3]. The current gold-standard for TCD CVR measurement is a 2-point breath hold or gas challenge, however this assumes a linear blood velocity response, neglecting plateaus at the upper and lower bounds of the CVR response [4, 5]. Notably, non-linear features of CVR may help distinguish different types of pathology [6, 7]. Although advanced alternatives have been developed for CVR mapping using MRI, the high cost and complexity of these setups hinder their application in clinical settings. In this regard, the development of a method to assess a dynamic range of CVR that is feasible and affordable for wide clinical use is warranted [8]. We present a novel breathing protocol in combination with TCD measurements for mapping non-linear CVR.

Dynamics of the cerebral blood flow (CBF) response to a novel ramped breathing protocol was assessed in 11 healthy participants (5F, aged 33±9 years). Blood flow velocity (CBFv) in the left MCA was measured continuously using a clinical TCD (Doppler-BoxX, DWL). CBFv changes were assessed in response to a ramp protocol consisting of 3 repetitions of 5 deep breaths, followed by 30s of regular breathing on medical air, 40s of 5% CO2 gas, and 40s of 10% CO2 gas. Data processing and analysis were performed using custom Python scripts. A rolling mean of the MCA velocity (MCAv) was applied across the pulsatile MCAv signal. The end-tidal peaks in the CO2 and O2 time-courses were selected automatically. To account for measurement delay, a bulk shift was applied to each PETCO2 trace to maximise its cross-correlation with the mean MCAv signal. MCAvmean was normalised relative to the mean MCAv during the baseline period (breathing air) to account for any variations in probe angle relative to the MCA. To characterise CVR, a 4-parameter sigmoid was fit to the MCAvmean vs. PETCO2 as shown in the equation in Fig 2 where a is the minimum blood velocity, b is the slope of the linear region, c describes the PETCO2 value for the inflection point, and d is the maximum velocity [9].

All 11 subjects successfully completed the ramp protocol. The timeseries of the MCAv, MCAvmean, CO2, and interpolated PETCO2 of a representative subject are presented in Fig 1. The normalised MCAvmean is plotted as a function of PETCO2 for the same subject in Fig 2 along with the sigmoidal fit. Across all subjects, the mean change in PETCO2 from hypocapnia (at the end of the deep breaths) to peak hypercapnia (at the end of the 10% CO2 period) was 28.9±4.7mmHg and the mean baseline PETCO2 was 26.0±5.1mmHg (t-stat>2.61 based on mean to max). The average increase in MCAvmean from hypocapnia to hypercapnia was 0.86±0.27% (t-stat>2.64) and the average baseline MCAvmean was 34.2±2.5cm/s. Changes in blood velocity and PETCO2 are highly correlated with all r-values>0.7 after cross correlation. The linear regression for all 11 subjects was statistically significant (p<<0.001, 0.7<r<0.9), however varying-degrees of plateauing of the blood velocity at the top and bottom of the PETCO2 range is apparent for all subjects. The sigmoidal fit was found to account for the non-linear features in the response with a better residual sum of squares for the sigmoidal fit compared to the linear fit (0.3±0.2 vs. 0.5±0.3 respectively), but refinement of the model bounds is needed to ensure physiologically accurate parameters.

A 4-parameter sigmoidal fit showed a promising model for characterising the full CVR response which may be particularly useful for the identification of vascular pathology. Future work includes developing a more sophisticated non-linear model of CVR and extending the use of a cost-effective ramp protocol and non-linear modelling in MRI.

Classification and Predictive Modeling

Methods Development ¹

Other Methods

Imaging Methods Other

Cerebral Metabolism and Hemodynamics ²

Acquisition

Blood

Cerebral Blood Flow

Data analysis

Design and Analysis

Experimental Design

Modeling

NORMAL HUMAN

ULTRASOUND