The identification of the tumour extent is a fundamental difficulty in brain tumour treatment planning and accurate measurement. Without ionising radiation, non-invasive magnetic resonance imaging (MRI) has emerged as a first-line diagnostic method for brain malignancies. Manually segmenting the extent of a brain tumour from 3D MRI volumes is a time-consuming job that heavily relies on the operator's knowledge. So, in this paper, we proposed a modified UNet structure that is based on residual networks that use shuffling periodically at encoder section of original UNet and sub-pixel convolution at decoder section. Sub-pixel convolution has the benefit over conventional resize convolution and it has extra parameters and thus stronger modelling capability at the same computing complexity and avoids de-convolution overlapping. The proposed UNet was tested on BraTS Challenge 2017 with high- grade glioma (HGG). The model was tested on BraTS 2017 and 2018 datasets. Tumour core (TC), whole tumour (WT), and enhancing core (EC) were the three major labels to be segmented (EC). The test results showed that the proposed UNet outperform the existing techniques.