Mixtures of lipids and detergents are known to form bicelles at certain parameter ranges, but many uncertainties remain concerning the details of the phase behaviour of these mixtures and the morphology of the formed lipid assemblies. Here we used nuclear magnetic resonance (NMR) diffusion data in combination with the multivariate processing method speedy component resolution (SCORE) to analyse mixtures of 1,2-dihexanoyl-sn-glycero-3-phosphocholine (DHPC) and 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) with the relative concentration q=[DMPC]/[DHPC]=0.5 at total lipid concentrations ranging from 15 to 300mM. With this approach we were able to resolve the heavily overlapping mixture spectra into component spectra and obtained reliable diffusion coefficients for lipid concentrations in the range 15 to 300mM, although at high concentrations (250-300mM), non-negativity constraints or overfactoring was required to successfully decompose the data. At 50-300mM total lipid concentration, the radii estimated from the diffusion coefficient of DMPC indicate assemblies of the appropriate bicelle size, although small size variations exist, while at lower concentrations the morphology appears to change to larger assemblies. Taken together, the results suggest that for q=0.5 DMPC/DHPC mixtures there is a relatively broad concentration range above 50mM where bicelles may reliably be assumed to adopt the 'classical' bicelle morphology. The study clearly demonstrates the usefulness of our approach for accurately determining physical properties of complex mixtures such as bicelles. Both reliable diffusion coefficients and chemical shifts can be derived from overlapping data. This should prove useful for analysing the behaviour of other, more complex, lipid mixtures.