3D ultrasound reconstruction allows physicians to explore a region of interest (ROI) in 3D while leveraging the advantages of 2D ultrasound imaging: simple, low cost and non-ionizing. It may assist many clinical tasks, such as organ measurement, procedure control or visualization of tissues difficult to interpret through 2D visualization. Recently, new deep learning techniques in the field of novel view synthesis, based on a continuous description of the 3D field, showed promising results in terms of 3D model estimation, robustness to noise and uncertainty, and efficiency. Inspired by these approaches, the objective of this work is to propose a 3D ultrasound reconstruction method based on neural implicit representations. Results on simulated and experimental data show the superiority of the proposed method compared to state-of-the-art voxel-based reconstruction.