GPU-based volume ray-casting can provide high performance for interactive medical visualization. The more samples we take along rays, i.e., a higher sampling rate, the more accurately we can represent the volume data, especially when the combined frequency of the volume and transfer function is high. However, this will reduce the rendering performance considerably because more samples mean more time-consuming memory access on GPU. In this paper, we propose an effective volume ray-casting algorithm which can perform more samplings within a ray segment using cubic B-spline. This can improve the sampling rate and offer high quality images without obvious performance degradation. Besides, our algorithm does not have to adjust anything else at all. This fact guarantees its flexibility and simplicity. We exploit the new programming interface CUDA to implement ray-casting rather than conventional fragment shader. Experimental results prove this method can be used as an effective medical visualization tool.