CardioBrain: a novel user-friendly software for assessing dynamic cerebral autoregulation — implementation validation and robustness to signal artefacts

OBJECTIVE: Cerebral autoregulation (CA) is a critical mechanism that protects the brain against ischemia and hyperperfusion by maintaining cerebral blood flow (CBF) relatively stable despite fluctuations in arterial blood pressure (ABP). Alterations in CA are linked to adverse outcomes, highlighting the need for reliable analysis methods. This study aimed to present CardioBrain, a free and user-friendly software for assessing dynamic CA (dCA) using transfer function analysis (TFA) of continuously recorded ABP and CBF velocity (CBFv) data. We also performed validation of the TFA implementation and evaluated software robustness to simulated signal artefacts and artefact filtering. Methods: The software calculates dCA metrics, including gain, phase, and coherence, across frequency bands (VLF, LF, HF). CardioBrain allows users to adjust signal-processing parameters following standards recommended by the Cerebrovascular Research Network (CARNet). Validation of the software's TFA implementation was performed by assessing agreement between CardioBrain and the TFA function provided on CARNet's website (tfa_car) using 18 ABP and CBFv signals. Robustness was assessed following the introduction of increasing levels of simulated signal artefacts (1-5%), with and without Hampel filtering. Results: Bland-Altman analysis showed excellent agreement between CardioBrain and tfa_car for coherence, gain (cm·s⁻¹·mmHg⁻¹), and phase (degrees), with mean differences (LoA) indicating negligible errors: VLFcoherence=-0.00(-0.000067 to 0.000054), VLFgain=-0.00(-0.000067 to 0.000066), VLFphase=-0.00(-0.000066 to 0.000053), LFcoherence=-0.00(-0.000063 to 0.000069), LFgain=-0.00(-0.000061 to 0.000041), LFphase =0.00(-0.000059 to 0.000069), HFcoherence=-0.00(-0.000058 to 0.000036), HFgain=-0.00(-0.00007 to 0.000063), HFphase = -0.00(0.000044 to 0.000029). Increasing artefact contamination, particularly from 2% onwards, resulted in greater bias and variability in TFA estimates, which were generally attenuated by Hampel filtering. Conclusion: This study presents CardioBrain as a user-friendly software tool for dCA assessment using TFA and demonstrates excellent agreement with an established TFA implementation. Robustness analyses demonstrate the impact of signal artefacts on TFA estimates and the potential of Hampel filtering to mitigate these effects.