Jul 2026· International Journal of Engineering Science and Technology· 0 citations
TL;DR
An explainable artificial intelligence framework for brain tumor classification using Gradient-weighted Class Activation Mapping (Grad-CAM), which generates class-specific heatmaps from a trained convolutional network and overlays them on MRI images to show the regions that most strongly influence a prediction.
Abstract
Deep neural networks can classify brain MRI images with high predictive capability, yet their internal decision processes are difficult to interpret. This limitation is especially important in medical imaging, where a prediction should be supported by evidence that can be examined by clinicians and researchers. This paper presents an explainable artificial intelligence framework for brain tumor classification using Gradient-weighted Class Activation Mapping (Grad-CAM). The proposed framework generates class-specific heatmaps from a trained convolutional network and overlays them on MRI images to show the regions that most strongly influence a prediction. The paper explains the mathematical intuition of Grad-CAM, provides a step-by-step implementation and validation protocol, and distinguishes visually attractive heatmaps from clinically meaningful explanations. Particular attention is given to layer selection, normalization, faithfulness, localization quality, uncertainty, and expert review. The framework also identifies common failure modes, including attention to text markers, skull boundaries, background artifacts, or preprocessing traces. A structured reporting checklist is proposed so that explanation results are assessed alongside classification performance. The paper concludes that Grad-CAM can improve transparency and error analysis, but it should be treated as supporting evidence rather than proof of causal reasoning or clinical correctness.
The integration of Artificial Intelligence in medical imaging offers significant potential for enhancing diagnostic speed and accuracy, yet the “black-box” nature of deep learning models remains a barrier to clinical adoption. This paper presents a robust framework for brain tumor detection utilizing a Convolutional Neural Network (CNN) integrated with Gradient-weighted Class Activation Mapping (Grad-CAM) to provide anatomical explainability. The proposed model utilizes a three-tier convolutional architecture optimized with Adam and binary cross-entropy loss. Experimental results on a brain MRI dataset demonstrate a training accuracy of 84.47% and a validation accuracy of 82.93%, indicating strong generalization capabilities. To bridge the gap between performance and interpretability, Grad-CAM was employed to generate visual heatmaps that localize the pathological regions of interest. These visualizations confirm that the model’s classifications are grounded in relevant tissue irregularities rather than image artifacts. This study demonstrates how explainable deep learning can serve as a transparent decision-support tool, fostering trust and providing verifiable diagnostic assistance for medical professionals.
Sungho Kim, Khairul Anam, Abdullah Al Mamu et al.· International Journal of Gri...· 0 citations
Correctly sorting brain tumors captured through Magnetic Resonance Imaging (MRI) plays a vital role in timely diagnosis and sound clinical decision-making, since a delayed or wrong call can seriously harm patient outcomes. This work introduces AMC-NeuroDx, an upgraded deep-learning pipeline that pits a Custom Convolutional Neural Network (CNN) built from the ground up against a ResNet50 transfer-learning model on a four-way brain tumor classification task (Glioma, Meningioma, Pituitary Tumor, and No Tumor), drawing on 7,223 MRI scans taken from the Kaggle Brain Tumor MRI dataset. ResNet50 follows a two-stage routine—training with a frozen base before full fine-tuning—and reaches 93.16% validation accuracy after only 20 epochs, whereas the Custom CNN needs 50 epochs to hit 90.70%. The framework also embeds Grad-CAM so that spatial heatmaps can show which brain regions drove each prediction, meeting the clinical demand for transparency. On top of this, a Streamlit interface lets users upload MRI scans in real time and automatically produces patient-specific PDF clinical reports holding the diagnosis, confidence scores, Grad-CAM overlays, and recommended next steps. The results obtained here show that transfer learning yields higher accuracy and quicker convergence than building a CNN from the ground up, and together with the built-in explainability and automated report writing, this makes AMC-NeuroDx a practical candidate for real clinical settings.
Kavya R, Mala M· International Research Journ...· 0 citations
Magnetic resonance imaging is widely used for the examination of brain abnormalities because it provides detailed soft-tissue information without ionizing radiation. Nevertheless, manual interpretation of large numbers of MRI slices is time-consuming and may be affected by inter-observer variation. This paper presents a structured deep learning framework for multi-class brain tumor classification using the EfficientNet family of convolutional neural networks. The framework emphasizes consistent preprocessing, transfer learning, class-balanced augmentation, careful validation, and clinically meaningful performance reporting. Rather than treating classification as an isolated model-training task, the proposed approach connects data quality, model calibration, error analysis, and reproducibility. The paper also explains why EfficientNet is suitable for medical image classification: its compound scaling strategy balances network depth, width, and input resolution, enabling strong feature learning with comparatively efficient computation. A complete experimental protocol is described for separating training, validation, and test data at the patient level; controlling information leakage; selecting evaluation metrics; and comparing EfficientNet variants with conventional convolutional baselines. The resulting framework can support reliable thesis-level experimentation and can later be extended with visual explanation methods such as Grad-CAM. The study concludes that model efficiency alone is insufficient; dependable brain tumor classification requires disciplined data handling, transparent reporting, and external validation.
Ajay Khatri, Sanmati Jain· International Journal of Eng...· 0 citations
Brain tumor classification plays an important role in early diagnosis and treatment planning. The current study aims to evaluate and compare the performance of five pre-trained convolutional neural network (CNN) models, namely VGG16, VGG19, MobileNet, Xception, and InceptionV3 using magnetic resonance imaging (MRI) images categorized into glioma, meningioma, pituitary tumor, and no tumor classes. To enhance model performance and address class imbalance, transfer learning and data augmentation techniques were employed. To boost model interpretability, heatmaps of important areas in tumor classification were produced through gradient-weighted class activation mapping (Grad-CAM). MobileNet was the most accurate with 97% and was more precise and more sensitive. The Grad-CAM visualizations showed the models were attending to clinically relevant features, which increased the interpretability. This comparative study demonstrates the effectiveness of the integration of explainable artificial intelligence (XAI) in deep learning pipelines for reliable brain tumor diagnosis.
Md. Firoz Hasan, Md. Awal Hadi, S. Nasrin et al.· IAES International Journal o...· 0 citations
A brain tumor classification system integrated with Explainable Artificial Intelligence (XAI) was developed using MRI images and demonstrated effective classification performance and improved interpretability, making it suitable for automated brain tumor diagnosis.
T. H. Stephen, A. Oke, A. S. Falohun et al.· LAUTECH Journal of Engineeri...· 0 citations
BACKGROUND AND OBJECTIVE
Explainable artificial intelligence is essential for clinical adoption of deep learning models in prostate magnetic resonance imaging. Although ensemble learning can improve robustness, its impact on explanation stability, spatial consistency, and clinical interpretability remains insufficiently quantified. This study aims to evaluate post-hoc interpretability methods across both single-model and ensemble configurations, and to examine whether ensemble-based explanations provide more reliable and clinically meaningful insights than single-model explanations. Critically, this work treats interpretability as a measurable property rather than a purely qualitative visualization.
METHODS
Convolutional neural networks and bagging-based ensemble models (five VGG16-based classifiers trained on bootstrap samples with replacement, aggregated by soft averaging) were trained on the public PROSTATEx dataset using T2-weighted and apparent diffusion coefficient images. Visual explanations were generated using Gradient-weighted Class Activation Mapping (Grad-CAM) and saliency maps. Lesion localization was evaluated using centroid distance and Dice similarity coefficient with expert-annotated lesion masks. An agreement metric was introduced to quantify spatial consistency between attribution methods and its relationship with prediction reliability.
RESULTS
The baseline classifier achieved an area under the curve of 0.84, with sensitivity of 0.81 and specificity of 0.86. Grad-CAM localized lesion centroids with higher precision on T2-weighted images (mean error 6.93 pixels) than apparent diffusion coefficient images (mean error 16.3 pixels). Combining saliency maps and Grad-CAM improved the mean Dice score from 0.42/0.45 (individual methods) to 0.52. Ensemble-based explanations were significantly smoother and less variable than individual classifier explanations (Mann-Whitney U, Levene test, all p<0.001). The agreement metric strongly separated correctly and incorrectly classified cases (Mann-Whitney U=84.5, p<0.001; point-biserial r=0.749; ROC-AUC =0.960).
CONCLUSIONS
These findings suggest that interpretability quality can be quantitatively assessed and improved through multi-method and ensemble-based analysis. The proposed agreement-driven framework enhances explanation robustness and supports reliable and transparent clinical decision support for prostate magnetic resonance imaging.
M. Gulum, M. Kantardzic, Christopher M. Trombley et al.· Computer Methods and Program...· 0 citations