Improving the quality of low-resolution Brain MRI images while preserving the crucial details is a challenging task. Many current methods like SRCNN and GAN-based methods like SRGAN fail to maintain the intricate details needed for accurate and error-free diagnosis of tumors. To address this issue, a deep learning approach augmented with attention-based channel and spatial networks is proposed. There is a Feature Reconstruction network (FRN) which aims to recover loss of information by recovering high-frequency features and upscaling the low-resolution image to the dimensions of the high-resolution image. There is also attention-based network that focuses on identifying the important regions like tumor regions. Once these regions are identified, targeted improvements are made to result in more accurate Brain MRI images. The incorporation of the attention mechanism in conjunction with a Feature Reconstruction Network (FRN) yielded encouraging outcomes. Evaluations using established metrics like Structural Similarity Index Measure (SSIM) and Perceptual Loss (based on VGG19) show that the model performs exceptionally well, achieving an SSIM score of 0.8061 and a Peak Signal-to-Noise Ratio (PSNR) of 76.70. This improvement directly allows the radiologists to detect tumors accurately while reducing the radiation exposure for patients.