Prostate cancer is the most commonly diagnosed cancer in men in North America. Transrectal ultrasound imaging (TRUS) is widely used for the evaluation of prostate disease. However, with conventional TRUS, diagnosticians must mentally integrate a series of two-dimensional (2D) images in order to develop an impression of the 3D anatomy, and the accurate estimation of prostate volume is difficult. We propose using 3D TRUS to overcome these problems. In this paper, we describe a 3D ultrasound imaging system and study its performance. The system consists of a conventional ultrasound machine, a microcomputer with a video frame grabber, and a custom-built assembly for rotating the probe. A typical scan of 200 2D B- mode images takes 13 seconds. These images can then be reconstructed into a 3D image, and the resulting 3D image can be interactively displayed using 3D visualization software. We also show that manual planimetry of 3D TRUS images can be used to estimate prostate volumes in vitro with 2.6% accuracy and 2.5% precision; and in vivo with 5.1% intra-observer variability and 11.4% inter-observer variability. Thus, 3D TRUS overcomes the limitations of 2D TRUS, and has great potential as a tool for the diagnosis and management of prostate disease.