Automated Segmentation and Classification of Tumors in MRI using Deep Convolutional Networks

Abstract

Magnetic Resonance Imaging (MRI) has become one of the most powerful diagnostic tools for identifying and characterizing tumors due to its superior soft tissue contrast and non-invasive nature. However, the manual delineation of tumors from MRI scans remains a time-consuming and error-prone process, heavily reliant on radiologists’ expertise and subject to significant inter-observer variability. To address these challenges, the present study focuses on the automated segmentation and classification of tumors in MRI images using Deep Convolutional Neural Networks (DCNNs). The proposed model integrates a multi-scale convolutional architecture capable of capturing both local and contextual spatial features, enabling precise boundary detection and reliable tissue differentiation. The research employs a hybrid approach that combines encoder-decoder structures, such as U-Net variants, with transfer learning from pre-trained deep networks like ResNet-50 and VGG-16 to enhance feature extraction efficiency. The dataset comprises annotated MRI scans obtained from open medical repositories, carefully pre-processed through normalization, noise reduction, and contrast enhancement. Data augmentation techniques, including rotation, flipping, and elastic deformation, are applied to overcome class imbalance and improve generalization. The network is trained using a categorical cross-entropy loss function optimized with the Adam optimizer, while segmentation performance is evaluated using Dice Similarity Coefficient (DSC), Intersection over Union (IoU), Precision, Recall, and F1-score metrics. Experimental results demonstrate that the proposed DCNN achieves superior segmentation accuracy and classification reliability compared to conventional thresholding and machine-learning-based methods. The average Dice coefficient exceeds 0.93, indicating excellent overlap between predicted and ground-truth tumor regions. Moreover, the network effectively distinguishes between benign and malignant tumors with an overall classification accuracy above 96%. Visual inspection confirms that the model preserves fine structural details, minimizing false positives in surrounding healthy tissues. The findings of this research highlight the transformative potential of deep learning frameworks in medical image analysis. Automated tumor segmentation and classification not only accelerate the diagnostic workflow but also provide consistent, objective, and reproducible outcomes that support clinical decision-making. This study underscores the importance of integrating artificial intelligence into radiological practices to enhance diagnostic precision, reduce workload on clinicians, and pave the way for personalized treatment planning in oncology.