The marching cubes (MC) is an effective surface rendering method in three-dimensional visualization for medical image sequence. However, most existing MC algorithms are slow and inefficient because they have to process all the cubes for isosurface extraction. An efficient MC algorithm is proposed by combing the case lookup table and a novel adjacent lookup sub-table to exclude unrelated empty cubes. By explicitly building the fix-length adjacent lookup sub-table that is independent on image sequences, the volumes elements are constrained to those only intersect with the isosurface. Both execution time and temporary storage space are further reduced by incorporating the heap data structure in algorithmic implementation. Experimental results on human head, chest foot CT data sets by using the visualization toolkit package show that the traversal cubes decreases by 95%, and the reconstruction time decreases by 20% without any loss of reconstruction quality. Therefore, the proposed method can remarkably speed up the rendering time and be easily integrated into 3D visualization for clinical application.