We demonstrated the first-ever use of a wash-in multi-breath SV imaging technique on consciously breathing human subjects [1]. The multi-breath technique made it possible to administer a series of identical HP gas breaths at a specified tidal volume while imaging the subject after each breath at the equivalent point in the respiratory cycle. This technique allows us to measure regional ventilation based on signal buildup over the series of images and then compute alveolar oxygen tension and apparent diffusion coefficient at a final longer breath-hold [2], resulting in maps of the efficiency of air replacement during tidal breathing (SV), alveolar oxygen tension (P_AO₂), and a localized measure of airspace size and connectivity (ADC). It was shown that the multi-breath Hybrid HP MRI approach opens up an innovative paradigm for evaluating pulmonary function and structure—one optimized to the intrinsic physiology of human lung function [3]. To emphasize the power of this new methodology, a longitudinal human study was performed to determine the ability of quantitative multi-breath hybrid hyperpolarized ³He MRI to predict lung function decline (rapid change in FEV₁ in a follow-up pulmonary function test). The multi-breath technique became possible with the use of gas delivery system presented in [4], in order to administer a constant volume of the imaging gas for six time points considered for SV (breath-hold~1.5s) and the seventh breathing time-point for the ADC and P_AO₂ imaging (breath-hold,12s), refer to Figure 1A-B. FiO₂ and inspiration/expiration duration were kept constant for each breath. SV, P_AO₂ and ADC maps were constructed via a previously described iterative fitting model [2]. End-inspiratory slice-selective (ST=25mm) images (6.25×6.25mm²) were acquired covering the entire lung volume on a 1.5-T Sonata MRI scanner (Siemens Healthcare) using an 8-channel phase array chest coil (Stark Contrast), using an accelerated (GRAPPA) fast low angle shot pulse sequence with TR/TE of 6.8/3.2ms and nominal flip angle=5˚ for ventilation and P_AO₂ images and an interleaved diffusion-weighted sequence with TR/TE of 9.1/5.9ms for ADC. Baseline and follow-up (average 1.4 years) respiratory assessments consisted of a pulmonary function test (PFT), six-minute walk test (6MWT), St. George’s Respiratory Questionnaire (SGRQ), and multibreath hybrid HP ³He MRI (SV, P_AO₂, and ADC). Lungs were segmented into twelve isotropic regions of interest (ROIs), and the ROI’s means and variances of each parameter were calculated. In smokers, a more than 75 ml/year drop in FEV₁ was defined as rapid lung function decline status [5]. Predictors of this decline were tested from explanatory variables of baseline imaging markers (ROI’s means and heterogeneity), demographics, pack-years, SGRQ, 6MWT and PFT. Covariates were included in a multivariate stepwise logistic regression model to predict decline status with leave-one-out cross validation. Receivers operating characteristic (ROC) curves were compared between different prediction models. Not all subjects were able to perform the whole imaging protocol or the signal to noise (SNR) was not enough for fitting to P_AO₂ or ADC (average SNR=100 required). In these cases, missing data was replaced with a model-based maximum likelihood imputation. The experiment was performed on 46 subjects, divided into three groups of 9 healthy nonsmokers (5 males, 4 females, 41-63 years old), 28 asymptomatic smokers (18 males, 10 females, 40-63 years old, 28 pack-years) and 9 COPD subjects (8 males, 1 female, 45-69 years old). Among smokers 18.0% (5/28) showed a rapid decline in their follow-up FEV₁. Two models were established, one based on non-imaging markers and one with additional imaging data. In non-imaging regression model, smokers with rapid decline showed a higher SGRQ score (OR=1.05,95%CI,1.01-1.10) and lowered DLCO/VA (OR=0.52, 95%CI, 0.26-0.98). Baseline residual volume (RV) entered the prediction model as a confounding variable (OR=1.69,95%CI,0.65-4.81). In imaging-based prediction model of decline status, higher SV heterogeneity (OR=1.75,95%CI,1.32-2.34), P_AO₂ heterogeneity (OR=3.05, 95%CI, 2.54-5.89) and mean ADC (OR=2.17,95%CI,1.84-2.58) collected from the twelve local isotropic ROIs for each subject were found to be significant predictors. ROC curves for non-imaging and imaging predictors of rapid FEV₁ decline are shown in Figure 2A-B. A 21% increase was achieved in the area under the curve (AUC) of the ROC plots, when imaging data was added to the prediction model. More importantly, a ~50% increase in the sensitivity (58% to 88%) was observed in the prediction of future decline in the smokers in this study. Smokers who eventually suffered from lung function decline were aware of their symptoms (significantly higher baseline SGRQ score), though most baseline standard global measures were normal. It is evident that the underlying mechanisms that lead to later global changes are detectable through regional and functional imaging using HP gas MRI.