There is limited understanding about how heart rate (HR) influences the blood-oxygen level dependent (BOLD) signal. While the mechanism by which respiration induces fluctuation in the BOLD signal is relatively well understood, the mechanisms regarding the HR remains unclear. The application of canonical cardiac response function (CRF), or subject-specific CRF, is an effective method for creating nuisance regressors, which can be used to remove cardiac-induced fluctuations in the BOLD signal. However, the relationship between physiological parameters and the characteristics of the CRF has not been systematically investigated.
In the present investigation, we studied the relationship between the variations in mean HR and the shape of the cardiac response function in 84 healthy subjects with a wide range of HR lying between 47 and 97 beats per minute (bpm). Three groups (n = 28) were created based on the subject's mean HR. We demonstrated that the HR plays an important role in determining the shape of the CRFs. We also observed that the canonical CRF explains more variance in subjects with a slow HR, than in subjects exhibiting faster HR. We found that the amount of explained variance significantly increased in each group when a group-specific CRF was used.
In a further analysis, we found two forms of a CRF, which explain a considerable amount of variance in subjects with a mean HR below and above 68 bpm. The shape of the CRF in subjects below 68 bpm is characterized by a shape similar to the canonical CRF, while in subjects with a HR above 68 bpm a well-defined second maximum was identified around 17 s. Thus, in the present study, we provide evidence for the necessity to use mean HR-based CRFs, rather than one canonical CRF, in order to optimally describe the interaction between BOLD and HR signal in subjects with varying heart rates.