<p>The average life span of the patient is shortened because of the deadliest and most destructive disease called as a brain malignancy. The chance of survival of the patient is limited because of the misdiagnosis of brain tumors in humans. The urgent need for an effective diagnostic method is increased because of the rising prevalence of brain tumors in humans. One of the most commonly utilized tools for tumor analysis is called as Magnetic Resonance Imaging (MRI). The manual segmentation is the toughest process because of the vast data produced by the MRI. High accuracy in brain tumor detection is not attained by conventional approaches like ResNet-50, Inception V3 Convolutional Neural Networks, VGG16, and which greatly impact the health condition of the patient. Motivated by these factors, this research proposed a deep learning model to recognize brain cancer which aids in ensuring timely treatment for enhancing the patient’s life span. The proposed solution is more important for adopting efficient treatment approaches to potentially improve patient outcomes. It is also used for effectively managing the complications associated with the brain tumor to enhance the patient’s life style. From the diverse sources, the brain images are initially collected. Subsequently, from the input images extraction of shape and texture takes place and is concatenated with each other. After that, the Adaptive Trans-Recurrent Residual Convolutional Neural Network based on Unet (AT-R2Unet) is used for obtaining the segmented images that help to recognize the brain tumor. The temporal dependencies of the input images are captured by the transformer and recurrent structure in the AT-R2Unet which helps to boost the segmentation accuracy. In addition, the AT-R2Unet model attains precise segmentation results because of the residual connections available in the structure that prevent the issues of vanishing gradient in the segmentation process. Amplified Dollmaker Optimization (ADO) tunes the attributes of the AT-R2Unet to enhance its performance. The visibility of the tumor features is enhanced because of the segmentation and the location of the brain cancer precisely identified by this model. The segmented images and the concatenated features are given to the proposed Hybrid (1D/2D) Convolution-based Efficient Attention Network (HC-EANet) after the segmentation for detecting the presence of brain malignancy. The HC-EANet plays the most significant part in the detection of brain cancer as it captures the temporal as well as spatial features. The HC-EANet models only concentrate on the particular region of the segmented images and provide a rich feature representation that helps to identify the presence of brain tumors in humans within a minimum time. This mechanism helps to increase the chances of recovery by precisely diagnosing the brain lump. The superiority of the model is proved by the validation of the proposed model with the existing models. The reported results from the experiment demonstrated the designed framework achieved 98.16% of accuracy in brain tumor detection.</p>