Purpose: Lung pO 2 mapping with 3 He MRI assumes that the sources of signal decay with time during a breath-hold are radiofrequency depolarization and oxygen-dependent T 1 relaxation, but the method is sensitive to other sources of spatiotemporal signal change such as diffusion. The purpose of this work was to assess the use of 3 He pO 2 mapping in patients with chronic obstructive pulmonary disease. Methods: Ten patients with moderate to severe chronic obstructive pulmonary disease were scanned with a 3D single breath-hold pO 2 mapping sequence. Results: Images showed signal increasing over time in some lung regions due to delayed ventilation during breath-hold. Regions of physically unrealistic negative pO 2 values were seen in all patients, and regional mean pO 2 values of Ă0.3 bar were measured in the two patients most affected by delayed ventilation (where mean time to signal onset was 3-4 s). Conclusions: Movement of gas within the lungs during breath-hold causes regional changes in signal over time that are not related to oxygen concentration, leading to erroneous pO 2 measurements using the linear oxygen-dependent signal decay model. These spatio-temporal sources of signal change cannot be reliably separated at present, making pO 2 mapping using this methodology unreliable in chronic obstructive pulmonary disease patients with significant bullous emphysema or delayed ventilation. Magn Reson Med 71:1172-1178, 2014. V C 2013 Wiley Periodicals, Inc.Key words: hyperpolarized 3 He; partial pressure of oxygen; chronic obstructive pulmonary disease; delayed-ventilation; diffusionThe partial pressure of oxygen (pO 2 ) in the lung airspaces depends on ventilation and perfusion matching and is a regional indicator of gas exchange. Hyperpolarized 3 He MRI pO 2 mapping has been shown to provide quantitative, regional measures of pO 2 in healthy volunteers (HVs) (1-4), which are physiologically plausible. The method assumes that all signal depletion during a breath-hold is due to radiofrequency (RF) depolarization and oxygen-dependent T 1 effects (5) and these effects can be separated to calculate the regional pO 2 using typically a linear (1), or exponential (6), oxygen-dependent signal decay model.The method is, however, also sensitive to other sources of temporal signal change during the breath-hold acquisition time. pO 2 mapping methods by definition require time resolved images to be separated by a delay time Dt, which is of the same order of magnitude as the 3 He T 1 in air % 20 s (1,5). Hence, inter-frame delays of 2-5 s are typically used to measure the pO 2 with 3 He. On this time scale, inter-slice diffusion of gas with different polarization has been shown to be a source of significant error in slice selective 2D pO 2 acquisitions leading to an underestimation of pO 2 values (4). This can be mitigated by using a 3D sequence (4) where the whole lung experiences the same RF history. Nevertheless, inter-pixel diffusion is still a source of error in 3D pO 2 mapping if the pixel size Dz, is such that, where D i...