In the paper we show that rotational invariance can be improved in a neural network based EIT reconstruction approach by a suitably chosen permutation of the input data. The input space is partitioned to non-overlapping sectors, and the input signal is permuted so that it lies in one sector independent of the original rotation angle. We demonstrate the advantages of the method with computer simulations. The proposed approach yields better results in the inverse problem, and allows use of smaller networks with fewer training samples.