Simultaneous multi-slab (SMSlab) technique is a 3D acquisition method that can achieve optimal signal-to-noise ratio (SNR) efficiency for high-resolution diffusion-weighted imaging (DWI) or functional MRI (fMRI). However, boundary artifacts may restrain its application. Nonlinear inversion for slab profile encoding (NPEN) has been proposed for its correction, which needs long computation time. In this study, we propose to use a convolutional network for boundary artifacts correction. It can solve the problem in a short time and improve the signal-to-noise ratio (SNR), which is of great meaning for high-resolution whole-brain DWI and fMRI.