Three-dimensional (3D) imaging allows acquisition of morphological traits, surface and volume data. Focusing on a selected section of an animal, it is supposed that regular acquisitions of such images allowed to follow, for example, fat deposition or gastro-intestinal changes. This technology was applied to study abdominal volume (AV) changes of dairy cows, which was considered to be a good estimation of rumen contents (RC). It aims to study the links between these changes and feed efficiency (i.e. production per unit of dry matter ingested). To assess RC variations, the rumen was manually emptied (study 1) or artificially filled (study 2) on 17 and 18 Holstein cows, respectively. In study 3, 55 Holstein cows were fed individually successively 2 diets differing in energy and fibres contents. Feed efficiency of cows was determined on this latest diet. Retention of feed bolus was defined as the ratio between AV and feed intake. Including individual effect in the analysis indicated that the linear regression of the RC was accurate with errors of 1.6 to 2 (studies 1 & 2), whereas RC prediction using partial least squares regression was not so accurate (7.1% error). The 25% most efficient cows in study 3 had a 1.18 times retention higher than the 25% less efficient. These results suggested a longer retention of the food bolus and a higher digestibility. Preliminary results indicated that 3D-imaging opens opportunities to study rumen contents, which is of interest to study feed efficiency and select animals for a lower environment impact.