In this paper, we present an approach for getting an initial pose to use in a 2D/3D registration process for computer-assisted spine surgery. This is an iterative process that requires an initial pose close to the actual final pose. When using a proper initial pose, we get registrations within two millimeters of accuracy. Consequently, we developed a fully connected neural network (FCNN), which predicts the pose of a specific 2D image within an acceptable range. Therefore, we can use this result as the initial pose for the registration process. However, the inability of the FCNN for learning spatial attributes, and the decrease of the resolution of the images before inserting them in the FCNN, make the variance of the prediction large enough to make some of the predictions entirely out of the acceptable range. Additionally, new researches in deep learning field have shown that convolutional neural networks (CNNs) offer high advantages when the inputs of the net are images. We consider that using CNNs can help to improve our results, generalizing the system for a greater variety of inputs, and facilitating the integration with our current workflow. Then we present an outline for a CNN for our application, and some further steps we need to complete to achieve this implementation.