Brain tumors are unusual cellular growths within brain tissue and are not always able to seen through standard imaging techniques. This study focuses on brain MR images to investigate a method aimed at enhancing the visibility of tumor-affected areas. We proposed a novel method for the detection of brain tumors from MR images. The heart of our approach lies in the integration of RCNN, allowing for accurate tumor detection within the identified ROIs. Subsequently, a logistic regression is employed for distinguishing between images with and without tumors; our experimental findings showcase the efficacy of the proposed method, with a sensitivity of 0.9599, a specificity of 0.9631, and an overall accuracy of 96.15%. These findings underscore the potential of our approach in automating the detection and classification of brain tumor from MR images, which holds significant promise for streamlining medical diagnoses and improving patient care. A collection of 600 images has been selectively sampled from the Cheng dataset. Within this set, 287 images depict cases of brain tumors, while the remaining 313 images present normal brain scans.