To improve the imaging depth of high-resolution spectral-domain optical coherence tomography (OCT) in the live zebrafish brain, this study uses an electrically tunable lens. This lens enables precise axial movement of the focal point within the sample arm. OCT images are acquired at different focal positions, thereby overcoming the depth-of-focus limitation of single imaging.In addition, a SwinFusion image fusion network is employed for intelligent perception, extraction, and reconstruction of clear regions. This approach achieves high-quality image fusion.Experimental results show that, after dynamic focusing, the imaging depth increases from 180 μm (single imaging) to 1.26 mm. This depth allows whole-brain imaging of zebrafish. For the SwinFusion method, the edge strength, peak signal-to-noise ratio, information entropy, and standard deviation are 81.351, 13.633, 6.233, and 51.093, respectively. These values indicate higher imaging quality than the Gabor transform method and the MF-CNN method. In the reconstructed three-dimensional model, different brain regions—including the telencephalon, diencephalon, midbrain, and medulla oblongata—can be clearly distinguished. The study indicates that the combination of dynamic focusing and artificial intelligence image fusion can effectively resolve the limited imaging depth problem of OCT in live zebrafish brain imaging.