ResNet50-based Deep Learning model for accurate brain tumor detection in MRI scans

•Model's test accuracy is 97.35 %.•High accuracy aids early tumor detection.•Assists radiologists, reduces workload.•Robust and generalizable to diverse scenarios.•Paves way for future research directions. Utilizing the ResNet50 architecture with tailored modifications, this study presents a revolutionary deep learning framework for brain tumor diagnosis from MRI scans. The primary contribution consists of optimizing the customized layers and fine-tuning the pre-trained ResNet50 model to improve diagnosis accuracy and robustness in real-world medical circumstances. The model was developed using a balanced dataset of 2,577 MRI images that were divided evenly between patients with tumors and healthy instances. In order to improve the model's generalizability across various clinical circumstances, extensive data augmentation approaches were used. With good precision, recall, and F1-scores, the suggested method had a test accuracy of 97.35 %, demonstrating its efficacy in identifying tumor instances. These findings highlight the model's potential as a trustworthy diagnostic aid that may lighten radiologists' diagnostic burden and improve decision-making by providing a second opinion. This technique enhances the accuracy of brain tumor identification by fusing cutting-edge deep learning with medical imaging. It also lays the groundwork for earlier diagnosis and more effective treatment planning. Subsequent endeavors will center on tackling the pragmatic obstacles associated with clinical implementation and enhancing the model's adaptability to various healthcare settings. [Display omitted]