Quantitatively assessing the tissue stiffness with the acoustic radiation force impulse imaging (ARFI) method has proved its effectiveness in clinical trials, such as the staging of liver cirrhosis and diagnosis of benign and malignant breast tumor. However, it has also been found that, the stability of its result is affected by many factors, such as measurement depth and tissue’s anisotropy. In this study, several modified methods for shear wave speed estimation were designed based on the existing Radon transformation (RT) method, and were compared using ultrasound radio-frequency data collected from a self-developed ARFI system. These RT based algorithms were classified to two types: (I) RT being performed on the displacement matrix with the axes of time and lateral location, and (II) RT being performed on the displacement matrix with the axes of time and depth. Type (I) algorithm attempts to find the best fitting trajectory of shear wave propagation in the lateral direction at a given depth, while Type (II) algorithm tries to directly find the exact time points when the shear wave front passes each lateral location in the whole measurement depth range. Experiments were performed on the soft tissue mimicking phantom and the ex vivo pork tissue sample. The reliability of repeated measurements and the computation time of these modified algorithms were compared. These results can help to find the most stable and time-saving algorithm of shear wave speed estimation, and improve the measurement reliability of ARFI in clinical trials.