INTRODUCTION: Following the introduction of SparseMRI [1], the combination of Compressed Sensing (CS) and parallel imaging has been of great interest to further accelerate MRI acquisitions [2-10]. With the exception of the autocalibrated GRAPPA-like approach [10], all these methods require accurate coil sensitivity estimations to achieve good quality reconstructions. Common approaches for this estimation are based on reference scans or fully sampled acquisition of the k-space center. However, the quality of the coil sensitivity maps can be affected by motion between the preacquired reference scans and the undersampled acquisition, whereas the acquisition of the central k-space (typically between 20% and 30%) can limit the maximum achievable undersampling factor. To overcome these problems, we propose an autocalibrated approach for the combination of CS and SENSE (SparseSENSE and its equivalents [2-5]) which does not require extra central k-space acquisition. This approach is based on the sequential estimation of the coil sensitivity maps and SparseSENSE reconstruction, from the same data set.