Generating dynamic carbon-dioxide from the respiratory-volume time series: A feasibility study using neural networks

In the context of fMRI, carbon dioxide (CO 2 ) is a well-known vasodilator that has been widely used to monitor and interrogate vascular physiology. Moreover, spontaneous fluctuations in end-tidal carbon dioxide (PETCO 2 ) reflects changes in arterial CO 2 and has been demonstrated as the largest physiological noise source in the low-frequency range of the resting-state fMRI (rs-fMRI) signal. Increasing appreciation for the role of CO 2 in fMRI has given rise to methods that use it for physiological denoising or estimating cerebrovascular reactivity. However, the majority of rs-fMRI studies do not involve CO 2 recordings, and most often only heart rate and respiration are recorded. While the intrinsic link between these latter metrics and CO 2 led to suggested possible analytical models, they have not been widely applied. In this proof-of-concept study, we propose a deep learning approach to reconstruct CO 2 and PETCO 2 data from respiration waveforms in the resting state. We demonstrate that the one-to-one mapping between respiration and CO 2 recordings can be well predicted using fully convolutional networks (FCNs), achieving a Pearson correlation coefficient (r) of 0.946 ± 0.056 with the ground truth CO 2 . Moreover, dynamic PETCO 2 can be successfully derived from the predicted CO 2 , achieving r of 0.512 ± 0.269 with the ground truth. Importantly, the FCN-based methods outperform previously proposed analytical methods. In addition, we provide guidelines for quality assurance of respiration recordings for the purposes of CO 2 prediction. Our results demonstrate that dynamic CO 2 can be obtained from respiration-volume using neural networks, complementing the still few reports in deep-learning of physiological fMRI signals, and paving the way for further research in deep-learning based bio-signal processing.