The acinus represents the functional unit of the mammalian lung. It is defined as the small tree of gas-exchanging airways, which is fed by the most distal purely conducting airway. Different hypotheses exist on how the fine structure of the acinus changes during ventilation and development. Since in classical 2-dimensional (2D) sections of the lung the borders of the acini are not detectable, every study of acini requires 3-dimensional (3D) datasets. As a basis for further studies of pulmonary acini we imaged rodent lungs as close to life as possible using phase contrast synchrotron radiation-based X-ray tomographic microscopy (SRXTM), and developed a protocol for the segmentation of the alveolar septa. The method is based on a combined multilevel filtering approach. Seeds are automatically defined for separate regions of tissue and airspace during each 2D filtering level and then given as input to a 3D random walk segmentation. Thus, the different types of artifacts present in the images are treated separately, taking into account the sample's structural complexity. The proposed procedure yields high quality 3D segmentations of acinar microstructure that can be used for a reliable morphological analysis.