The size and geometry of the prostate are known to be pivotal quantities used by clinicians to assess the condition of the gland during prostate cancer screening. As an alternative to palpation, an increasing number of methods for estimation of the above-mentioned quantities are based on using imagery data of prostate. The necessity to process large volumes of such data creates a need for automatic segmentation tools which would allow the estimation to be carried out with maximum accuracy and efficiency. In particular, the use of transrectal ultrasound (TRUS) imaging in prostate cancer screening seems to be becoming a standard clinical practice due to the high benefit-to-cost ratio of this imaging modality. Unfortunately, the segmentation of TRUS images is still hampered by relatively low contrast and reduced SNR of the images, thereby requiring the segmentation algorithms to incorporate prior knowledge about the geometry of the gland. In this paper, a novel approach to the problem of segmenting the TRUS images is described. The proposed approach is based on the concept of distribution tracking, which provides a unified framework for modeling and fusing image-related and morphological features of the prostate. Moreover, the same framework allows the segmentation to be regularized via using a new type of "weak" shape priors, which minimally bias the estimation procedure, while rendering the latter stable and robust.