OBJECTIVE Oxygen-enhanced MRI (O2-MRI) is frequently based on a block paradigm consisting of a series of consecutive T1-weighted scans acquired during alternating blocks with inhalation of room air and of pure oxygen. This design results in a complex signal-time course for each pixel, which displays the oxygen wash-in and wash-out processes and provides spatially resolved information about the lung function. The purpose of the present study was to optimize the signal-time-course analysis to extract (pixelwise) the maximum amount of information from the acquired data, and to introduce an appropriate cross-correlation approach for data sets containing the oxygen wash-in and wash-out periods. METHODS O2-MRI data of 11 healthy volunteers were acquired with a multislice inversion-recovery single-shot turbo-spin-echo sequence at 1.5 Tesla; lung and spleen were manually segmented on all 44 acquired slices. Six different model functions were pixelwise fitted to the data and compared using the Akaike information criterion. Four different reference functions were compared for cross-correlation analysis. RESULTS The optimal model function is a piecewise exponential function (median enhancement in lung/spleen: 16.3%/14.8%) with different time constants for wash-in (29.4 seconds/72.7 seconds) and wash-out (25.1 seconds/29.6 seconds). As a new parameter, it contains the delay between switching the gas supply and the onset of the signal change (4.8 seconds/24.5 seconds). Optimal cross-correlation results were obtained with a piecewise exponential reference function, which was temporally shifted to maximize the correlation, yielding median correlation coefficients of 0.694 and 0.878, median time delays of 7.5 seconds and 38.6 seconds, and median fractions of oxygen-activated pixels of 83.6% and 92.2% in the lung and the spleen, respectively. CONCLUSIONS It was demonstrated that the pixelwise assessment of O2-MRI data are optimally performed with piecewise exponential functions. Cross-correlation analysis with a piecewise exponential reference function results in significantly higher fractions of oxygen-activated pixels than with rectangular functions.