Detection of Brains Tumor from MRI Images Using Convolutional Neural Network VGG19 Model

Precise and automatic using imaging techniques that use magnetic resonance (MRI) to detect brain tumours remains a formidable challenge owing to the tumours’ unusual shapes, diverse sizes, and intricate locations. Current methodologies frequently encounter challenges related to inadequate segmentation precision or restricted generalizability in classification endeavours. Although deep learning demonstrates significant potential, several frameworks concentrate exclusively on either segmentation or classification, resulting in insufficient diagnostic insights. An integrated system is urgently required to concurrently detect and define tumours with high precision and clinical reliability. We provide a hybrid DL system in this research that integrates the VGG19 architecture with a U-Net-based segmentation model to tackle tumour localization and binary classification. The segmentation module effectively delineates tumour locations, whereas a distinct VGG19-based classifier differentiates between tumour and nontumor MRI slices. The suggested method achieves a perfect score on the F1 test and a classification accuracy of 1.00, surpassing benchmark models including InceptionV3, ResNet50, and EfficientNetB0. Segmentation results show that the model is successful with a Dice Similarity Coefficient of 0.8509 and an IoU score of 0.7411 accurate tumour border delineation. This work’s originality is in the incorporation of high-resolution spatial features from VGG19 into a cohesive dual-task framework, accompanied by a thorough assessment of both pixel-level and image-level tasks. This dual-stage architecture overcomes shortcomings in previous investigations by guaranteeing reliable tumour recognition and delineation, positioning it as a potential tool for AI-assisted diagnostics in neuro-oncology..