In MRI, subject physiology is commonly recorded with pulse oximetry (PPU), electrocardiography (ECG) and/or respiratory belt for various purposes, including de-noising of fMRI ¹ Setting up those peripheral devices prolongs preparation time and reduces patient comfort (direct skin contact/tight fit). Recently, magnetic field recording with NMR field probes was suggested as a potential alternative ². Placed above the upper chest of a subject lying in the bore, they can provide highly sensitive readouts of both cardiac and respiratory processes without direct contact to the subject. In this work, we explore physiology recording with field probes during echo planar imaging and demonstrate the equivalence to standard methods for fMRI data de-noising (RETROICOR).EPI time series (32 transversal slices, 1.5x1.5x1.5 mm, slice TR 62.5 ms, SENSE factor 4, 150 dynamics) of a freely breathing volunteer were recorded at 7T (Philips Achieva, 1 ch TX, 32 ch RX array, NOVA Medical). Magnetic field evolutions were recorded during image encoding using two 19F NMR field probes connected to a standalone spectrometer ³. Field measurements were taken every 5 ms (10 per slice). One probe (A) was suspended anterior to the volunteer’s chest, close to the heart. The second probe was placed posterior to the volunteer’s chest, in the table padding (B) (fig.1). No modifications to the imaging sequence were necessary. For comparison, standard devices - PPU and respiratory belt - were recorded simultaneously. The probe signals were corrected for phase accrual caused by image encoding gradients. Linear regression of the probe phases yielded one field value per probe acquisition. Slow field drifts due to gradient heat up were corrected by the subtraction of 3rd-order polynomials. Data quality after processing (fig 2a) is comparable to signals acquired without imaging. Cardiac and respiratory contributions were extracted through frequency selective filtering (0.6-15Hz and 0.05-0.7Hz), sequenced with a pattern matching algorithm (PhysIO toolbox ⁴). RETROICOR was used to translate them to physiology regressors (fig. 2). Three independent regressor sets were created with probe A, probe B and standard devices respectively. Each set was used to identify physiological noise in the data with general linear modelling (SPM12). The measured field changes induced by physiology were on the order of 50-100 nT while the imaging gradient strength at the probe position was of about 3mT (Fig.2). In probe A, the cardiac contribution is approximatively 3 times stronger than the respiratory one. Conversely, in probe B, respiratory contribution is about 2.5 times stronger than the cardiac one. Fig.3 compares the F-maps (p< 0.001, uncorrected) for the probes and the standard devices models (left side). On the right side of fig.3, probes are compared to standard devices individually. The accordance between probe A and the standard devices amounts to 74% (cardiac) and 87% (respiratory) commonly detected pixels. For probe B the values are 77% and 84% respectively. NMR Probe-based magnetometry performed concurrently with MR imaging proved to provide robust physiological readouts. We demonstrated its utility for active physiological noise reduction in an fMRI scenario and showed its equivalence to standard peripheral measures. These results pave the way towards fully integrated and truly remote physiology monitoring with improved patient comfort and reduced preparation time.