Volumetric analysis of brain ventricles is important to the study of normal and abnormal development of the central nervous system of mouse embryos. High-frequency ultrasound (HFU) is frequently used to image embryos because HFU is real-time, non-invasive, and provides fine-resolution images. However, manual segmentation of ventricles from 3D HFU volumes remains challenging and time consuming. In this study, in utero and in vivo volumetric ultrasound data were acquired from pregnant mice using a 5-element, 40-MHz annular array. An automatic segmentation algorithm based on active shape model (ASM) was developed to segment the brain ventricles of the embryos; ASM allows us to efficiently 'learn' from training data (i.e., manually segmented data). The algorithm was further enhanced by using detail-preserving reference shapes (also learned from training data) and region growing constrained by the reference shape. The hybrid algorithm was applied to three 12.5-day-old embryos. Results were qualitatively analyzed and compared with manual segmentation results in regions typically difficult to segment (e.g., thin brain ventricle connections). In addition, quantitative analysis using the Dice similarity coefficient (DSC) was used to compare the automatic segmentation results with manual segmentation. We obtained average DSC values of 0.848±0.015 for the brain ventricles and our method produced morphologically accurate results. Therefore, our method could streamline current HFU longitudinal studies of brain development that require manual segmentation.