Cancer is one of the non-contagious diseases that causes the most significant number of human deaths. The time it takes for cancer cells to be identified in a patient’s body significantly impacts how easily they can be treated. An automatic screening method using computer-aided diagnostic (CAD) is one way of early can- cer detection. This dissertation proposes several models that can be adopted as automatic screening methods. The first model is a convolutional-based single neural network built by adopting the Visual Geometry Group (VGG) module. The second model is an ensemble model based on a voting system built by combining three single networks from scratch by adopting three well-known mod- ules: VGG, Inception, and Residual Network modules. The last model is an ensemble model based on interconnected networks. Un- like the previous ensemble model, this model does not use a voting method in decision making but trains all three networks in an ex- tensive linked network to make a single final decision. Furthermore, the success of each model, also the benefits and drawbacks of it are presented.