The X-ray exposure of hepatic MDCT (Multidetector-Row Computed Tomography) images threatens human health. In this paper, the 3D adaptive median filter with local averaging is optimized to improve the image quality of low dose hepatic MDCT images. The modified strategy is designed according to the subtle texture of a hepatic tumor. Filtering experiments are based on phantom images with different dose levels and clinical hepatic ones with a certain dose level. Qualitative and quantitative evaluations show that the x-ray exposure can be reduced to 60% at least without compromising the quality of diagnostic images.