After their triumph in various classification, recognition and segmentation problems, deep learning and convolutional networks are now making great strides in different inverse problems of imaging. Magnetic resonance image (MRI) reconstruction is an important imaging inverse problem, where deep learning methodologies are starting to make impact. In this work we will develop a new Convolutional Neural Network (CNN) based variant for MRI reconstruction. The developed algorithm is based on the recently proposed deep cascaded CNN (DC-CNN) structure. In the original DC-CNN network, the regular data consistency (DC) layer acts as a periodic enforcer of data fidelity. Here, we introduce a novel DC layer structure which also calculates the projection of the intermediary image estimate onto the unobserved subspace of the Fourier domain. These intermediary innovation images are saved and reutilized in the final stage of the overall structure via skip connections. This enhanced cascaded deep network results in improved reconstruction performance when compared to not only the original DC-CNN structure but also another recent deep network approach, where similar number of parameters get utilized in the competing deep methods.