Simulation of medical ultrasound images using linear systems theory.

摘要

Medical Ultrasonography provides an accurate real-time assessment of organs and systems without exposing the patient to harmful ionizing radiation. These features and the cost effectiveness of the procedure have made ultrasonography one of the most widely used diagnostic medical imaging modalities. The practice of ultrasound is expanding into diverse areas of clinical medicine, and as a result, there is an increasing demand for education and training in the use of ultrasound.; This project addresses a novel ultrasonographic simulation algorithm that has the ability to generate ultrasonograms from any position and orientation, from either the surface of or inside a virtual body. The algorithm is able to take as inputs the spatial location and ultrasound characteristics of the tissues and in turn produces images similar to true ultrasonograms with a frame rate comparable to real medical ultrasound instruments.; Initially, the trainee determines the location and orientation of the virtual transducer. From the obtained information, the simulator retrieves the necessary anatomic data from a segmented Visible Human dataset to localize the organs and tissues in the image plane. Then, just like real instruments, the ultrasound beams are emitted from the virtual transducer. Waves reflected from interfaces and speckle noise from subwavelength scatterers are calculated based on the linear systems theory, tissue properties, and the physics of ultrasound. In the case of high amplitude reflections, the simulator also processes the reverberations. The results pass through filters for time gain compensation and motion blur. After correcting for brightness and contrast, the image is displayed on a screen as the ultrasonogram of the volume under investigation. The simulator also labels the various structures in order to clarify the images under examination.; Statistical analyses of the images confirmed the similarity between the simulated and the real ultrasonograms. Frequency domain analyses also demonstrated identical frequency patterns between the simulated and real images. Each frame took 186.8 milliseconds on average to generate on a 2.0 GHz Pentium 4 system with 1 GB random access memory, which is comparable to real ultrasound instruments.; This study demonstrates the feasibility of creating simulated ultrasonograms for interactive training and education.