Application of innovative image processing methods and AdaBound-SE-DenseNet to optimize the diagnosis performance of meningiomas and gliomas

•A novel multi-directional brain region extraction method (MDBRE) method is proposed to remove the disturbances of skulls and background.•A novel iterative gamma correction based on two peaks (TPGC) is proposed to narrow the illumination variances among brain images.•The advantages of DenseNet, SE blocks and AdaBound are combined to form an AD-SE-DenseNet that can contribute to improving the diagnosis performance.•More than 4800 original brain images are employed to diagnosis meningiomas, gliomas and normal for the first time.•The experimental results indicate that the proposed system can achieve the state-of-the-art performance.