Research Information System
vol.37, no.1, pp.1-18, 2026 (Peer-Reviewed Journal)
ABSTRACT:Background:Proper and interpretable brain tumor clas-sification is essential in making a successful clinical decision in neuro-oncology. Automated approaches are potentially promising, but a lack of trans-parency in decision-making is usually an obstacle to clinical implementation.Objective:The proposed research developed and evaluated a convolutionalneural network (CNN) model in the context of automatic brain tumor clas-sification using magnetic resonance images (MRI) with a particular focus ona high-performing model and visualizing predictions using Gradient-weightedClass Activation Mapping (Grad-CAM).Methods:It utilized a dataset of7,023 MRI scans as a sample, which was divided into glioma, meningioma, pi-tuitary tumors, and no-tumor. Preprocessing of the data was done by normaliz-ing and resizing, and stratifying into training, validation, and test subsets. Thesuggested CNN has been compared with the state-of-the-art transfer-learning ar-chitectures, such as VGG16, MobileNetV2, and DenseNet121.Results:Theproposed CNN had the highest predictive accuracy of 94.75%, precision of94.99%, recall of 94.75%, and an F1-score of 94.82%, and better than all thetransfer-learning baselines. Moreover, Grad-CAM visualizations have alwaysidenABSTRACT:Background:Proper and interpretable brain tumor clas-sification is essential in making a successful clinical decision in neuro-oncology. Automated approaches are potentially promising, but a lack of trans-parency in decision-making is usually an obstacle to clinical implementation.Objective:The proposed research developed and evaluated a convolutionalneural network (CNN) model in the context of automatic brain tumor clas-sification using magnetic resonance images (MRI) with a particular focus ona high-performing model and visualizing predictions using Gradient-weightedClass Activation Mapping (Grad-CAM).Methods:It utilized a dataset of7,023 MRI scans as a sample, which was divided into glioma, meningioma, pi-tuitary tumors, and no-tumor. Preprocessing of the data was done by normaliz-ing and resizing, and stratifying into training, validation, and test subsets. Thesuggested CNN has been compared with the state-of-the-art transfer-learning ar-chitectures, such as VGG16, MobileNetV2, and DenseNet121.Results:Theproposed CNN had the highest predictive accuracy of 94.75%, precision of94.99%, recall of 94.75%, and an F1-score of 94.82%, and better than all thetransfer-learning baselines. Moreover, Grad-CAM visualizations have alwaysidentified tumor-specific areas in the images, confirming the clinical plausibil-ity of the model decisions.Conclusions:These results highlight the possi-bility of high-performance CNN-based classification used in conjunction withexplainable AI to provide effective and high-quality diagnostic support thatis accurate, dependable, and explainable by clinicians. The future researchwill explore the concept of multi-modal MRI integration, 3D architecture, andprivacy-preserving deployment schemes in the context of real-life healthcareapplicationstified tumor-specific areas in the images, confirming the clinical plausibil-ity of the model decisions.Conclusions:These results highlight the possi-bility of high-performance CNN-based classification used in conjunction withexplainable AI to provide effective and high-quality diagnostic support thatis accurate, dependable, and explainable by clinicians. The future researchwill explore the concept of multi-modal MRI integration, 3D architecture, andprivacy-preserving deployment schemes in the context of real-life healthcareapplications