NeuroNet Diagnostics: Brain Tumor Detection and Segmentation Using Deep learning

The brain tumor is the most devastating disease, and the diagnosis should be on time and accurate. The traditional ways of diagnosis, for example, the interpretation of the data in MRI by radiologists, are known to take up a lot of time and are subjective. New trends in deep learning open new doors for auto-mated systems, thereby diagnosing and classifying brain tumors with more accuracy, which will facilitate earlier detection and less workloads for medical professionals. This article develops a novel framework in finding and segmenting brain tumors in its advanced preprocessing technique of incorporating images in the concept of deep learning. Adaptive bilateral filtering with the aid of VGG-19 CNN architecture is applied in improving the quality of the images with precision. This model takes advantage of the transfer learning; thus, refining the outputs based on the classification of either tumorous regions and non-tumor regions or even pixel-wise. It categorizes the tumour into glioma, meningioma, and pituitary types by using annotated MRI datasets with data augmentation that has assisted in overcoming sparse data challenges. Performance metrics namely F1-score, sensitivity, specificity, and accuracy are taken into account for the analysis of the model. The experiment results show an impressive accuracy of 98.91% -99.21%. The approach promises good clinical usage and provides a good tool for radiologists that can help them diagnose brain tumors more efficiently and accurately to improve patient care.