Probability Mass Function-Based Adaptive Median Filtering for Acoustic Radiation Force Impulse Imaging: A Feasibility Study

Among ultrasound images, the acoustic radiation force impulse (ARFI) image can provide information about the stiffness of the tissue using both a pushing beam and a detection beam. However, there is a problem in that the impulse noise is generated in the process of generating the ARFI image by calculating the displacement of the target. The impulse noise appearing in the ARFI image can be caused by various factors related to ultrasound image processing and the characteristics of the target tissue, and it can degrade image quality and the accuracy of stiffness measurements. The commonly used fixed median filter in the ARFI image can effectively eliminate the impulse noise but may introduce a blurring effect, depending on the kernel size. The adaptive median filter has the advantage of minimizing the impulse noise level while preserving the original information as much as possible, but the adaptive median filter has been generally used to remove the unipolar or bipolar type noise from the image. As a result, the effectiveness of removing the impulse noise with various amplitudes from the ARFI image is not sufficient. To figure out this problem, in this study, we propose the adaptive median filter method combined with probability mass function. In this method, in order to limit the various amplitudes of impulse noise as much as possible, the threshold cut-off level of impulse noise is determined by a probability mass function, and then adaptive median filter is used to effectively remove impulse noise with limited amplitude. The performance of the proposed method was evaluated by using a tissue mimicking phantom and a bovine eye. Therefore, the proposed technique is expected to be one of the useful methods to improve the overall quality and reliability of the ARFI images for clinical diagnosis and evaluation.