When considering oxygenation in most organs we generally think about O2 delivery, O2 metabolism and tissue PO2 levels.

However the lung is different. The lung is a gas exchange organ, so the primary function of the lung is not to utilize O2 for metabolic processes, but to transfer oxygen from the atmosphere to the blood and to transfer CO2 from the blood back to the atmosphere. So our metrics when considering oxygenation in the lung are different. To quantify oxygen transferred by the lung to the blood we need more than just measurements of alveolar ventilation (V). We also need to know the capillary perfusion (Q), and most importantly, how well these are matched (the V/Q ratio). Areas that are well perfused but poorly ventilated (“shunts”) or well ventilated but poorly perfused (“dead space”) will both be inefficient at gas exchange.

MRI of the lung presents some significant challenges, and traditionally, MRI has not been considered a suitable imaging method for measuring lung physiology. The presence of air/water interfaces results in local susceptibility gradients that significantly shorten the local T2* and result in low SNR. The entire lung is full of air/water interfaces, so we need to be selective in our choice of pulse sequence design to work with the very short T2* values. In addition the lung is in motion, and while short breatholds are manageable, prolonged breatholds are challenging for healthy volunteers, and often impossible in patients and can also perturb the local physiology.

For this presentation I will focus primarily on some novel methods for acquiring physiological measures of lung function using proton MRI that can be done on a clinical MRI scanner.

The first part of our challenge is to measure capillary perfusion in the lung. An established method for this in brain and other organs is arterial spin labeling (ASL). However, this traditionally uses fast gradient echo techniques with EPI or spiral k-space trajectories that are not suitable for lung imaging, so we need to resort to fast spin echo T2 imaging methods instead. Our aim is to quantify the amount of blood delivered to the lung capillaries in one cardiac cycle, so we need to gate the images to both the lung and the heart. In practice we use a short breath hold at our lung inflation level of choice (e.g, functional residual capacity or total lung capacity) to prevent lung motion, and we use ECG gating to determine the triggering and acquisition window based on the heart rate. As with other ASL techniques a pair of images are collected with and without an inversion tag, and the MRI signal difference between them is related to spins delivered via pulmonary perfusion. To quantify the MRI signal units in terms of actual ml of blood delivered per minute per ml of lung tissue we need a scaling factor related to water density (often obtained from a water phantom in the field of view), and additionally to correct for loss of the tagged signal due to T1 of blood and correct for any inhomogeneities in our receiver coils. The large conduit vessels in the subtracted images represent blood in transit and not blood yet delivered to the capillary bed, so these need to be excluded. Since the lung tends to sag under its own weight the density of lung is greater in the more dependent parts, and thus the capillary density will also be greater, so to allow for this we also need to correct the ASL images for the variations in lung density across the lung. What we end up with is a regional measure of density-normalized pulmonary perfusion.

Next we need to measure alveolar ventilation (i.e. the volume of fresh gas reaching the alveoli per minute). We need a MR-visible gaseous contrast agent, and for this we can actually use oxygen itself by switching the gas the subject is breathing from air to 100% O2. Over time the local tissue pO2 will rise to a new steady-state level. The key idea here is that the rate at which local pO2 approaches this new steady-state value depends on local specific ventilation, the ratio of fresh gas delivered with each breath to total gas in the alveoli. Dissolved O2 is paramagnetic and will tend to reduce the local T1 of surrounding water, increasing the MR signal in a T1-weighted image, so the approach to pO2 equilibrium as the subject continues to breath 100% O2 can be measured from the time course of the MR signal between two steady-states. Note that, the key measurement here is not the magnitude of the T1 shortening but the rate at which the signal changes. In the transition from breathing air to breathing 100% O2, areas of high ventilation will reach a new steady-state MR signal level faster. Conversely lung regions with lower ventilation will approach a new steady-state value more slowly. By alternating between air (21% O2) and 100% O2 we can build up wash-in and wash-out curves for O2. This rate at which regions achieve equilibrium actually gives us a direct measure of specific ventilation. To convert specific ventilation (the ratio of fresh gas delivered per breath to total gas volume) to alveolar ventilation (absolute volume of fresh gas delivered) we need a measure of the gas volume of the local alveoli. If we assume that lung is basically air and water, we can calculate the proportion of air if we know the proportion of water, and we can measure that from a proton density measurement. Some additional postprocessing is again needed to correct for inhomogeneities in the RF coil sensitivity profile. Images from different sequences are also co-registered and smoothed.

Once we have measures of both ventilation and perfusion we can then examine how well they are matched, and this V/Q ratio gives us a direct measure of gas transfer efficiency.