An attempt is made in this paper to diagnose brain-related diseases like sarcoma, fatal stroke disease, cerebral disease, and Alzheimer's disease by using saliency information from magnetic resonance and computed tomography source images. The saliency information for each source image is computed using guided and median filters. The obtained saliency maps are used for computing the weight maps of each source image by using image statistics. The obtained weights are used to fuse the approximate and detailed layers of the source images by using the weighted average fusion technique. The proposed algorithm is simulated in MATLAB for various benchmark data sets of brains taken from the brain atlas provided by Harvard Medical School, available at https://www.med.harvard.edu/aanlib. In order to test the efficacy of the novel method, comparative analysis is performed in terms of image quality assessment metrics like mean, mutual information, average gradient, standard deviation, spatial frequency, etc. From the analysis, this paper concludes that the proposed algorithm improved gradient information in the fused image by 35.7%, entropy by 5.7%, spatial frequency by 32.7%, edge strength by 14.5%, and minimized the information loss by 43.6%. Therefore, the novel method of weight map computation produces a detailed and noise-free image, which is helpful for better diagnosis in clinical applications.