The filtration through machine learning approaches can solves many problems in diagnosis of MRI brain. Various medical image applications are available in usage but they are offering limited solutions at diagnosis process. Therefore disease diagnosis performance is improved by proposed CNN (convolution neural networks)–GB (Gradient Boosting) and adaptive median filter (AMF) mechanism. In this work, brain-related disorders have been identified by using deep learning and machine learning technique. So brain-correlated information purpose real-time MRI medical images and “FIGSHARE”, “OASIS” datasets are collected. These are having more information of disease identification and classification. Various filters like Gaussian filter, median filters unable to remove the noise in medical images, when density of noise is more than 78% so that an advanced filters design is necessary for a fast and accurate brain disease diagnosis. In this research CNN–ML based adaptive median filter is designed for noise exclusion and effective diseases identification. PSNR, MSE, True positive rate and accuracy parameters can describe the application efficiency. MRI brain diseases investigation with CNN–AMF and GBML model has achieved 0.9863-accuracy and 0.9832-True Positive Rate, which is a better improvement. This implemented design and improved results are compete with current medical research applications, also useful for doctors, researchers and Diagnosis centers.