In medical image segmentation, deep learning-based network methods perform well, during the last few years, most of them are using U-Net, X-Net and SegNet. In this paper, we propose hybrid BConvLSTMSegX-Net with the concatenation of SegNet and X-Net. Instead of simply adding skip connections in SegX-Net, we take full advantage of BConvLSTM and also batch normalization is used to accelerate the network. Experimentation is carried out and the results are obtained for segmentation of brain tumor obtained Accuracy:0.98, Specificity:0.96, Precision:1.00 & F1- Score:0.95.