Brain Tumor Detection Using Deep Learning and Automated Report Generator (NeuroScanAI)

Abstract

Brain tumor is one of the most critical and life-threatening diseases, which requires early detection for effective treatment. We proudly present to you “NeuroScanAI”, a deep learning- based system for automated brain tumor detection using Magnetic Resonance Imaging (MRI) scans and generating a pdf format of medical report with basic patient details. The proposed approach utilizes Convolutional Neural Networks (CNNs) models with transfer learning technology on pre-trained data set such as MobileNetV2 to achieve high accuracy . The model attains an accuracy of approximately 83–84% across four tumor classes. To enhance interpretability, Gradient-weighted Class Activation Mapping (Grad-CAM) is deployed to highlight tumor regions which increases model predictions. Moreover, a web-based interface is developed to enable image upload, prediction visualization, and automated PDF report generation. “NeuroscanAI” demonstrates an effective combination of accuracy, explainability, and practical usability for medical image analysis and AI intergration.

Introduction

Brain tumors are life-threatening and difficult to diagnose early, with types like glioma, meningioma, and pituitary tumors requiring different treatments. MRI is commonly used for detection, but manual analysis is slow, costly, and prone to human error. To improve this, deep learning—especially CNNs and transfer learning—has been widely adopted because it can automatically learn features from MRI images and improve accuracy over traditional methods.

However, deep learning models often lack interpretability, which limits their use in healthcare. To address this, explainable AI techniques like Grad-CAM are used to highlight tumor-affected regions and improve trust in predictions. This study proposes NeuroScanAI, an end-to-end system that detects brain tumors, provides visual explanations, and generates automated diagnostic reports.

The system uses a Kaggle MRI dataset of about 3,000 images across four classes (glioma, meningioma, pituitary, and no tumor), split into training, validation, and testing sets (70/15/15). Images are preprocessed through resizing, normalization, and augmentation to improve model robustness. The model is based on MobileNetV2 with transfer learning, trained using the Adam optimizer, achieving around 83–84% accuracy.

Grad-CAM is integrated to visualize important regions influencing predictions, improving interpretability for clinicians. The system is implemented in Python using TensorFlow and other libraries, with a web interface (Gradio/Streamlit) that allows users to upload MRI scans, get predictions, view heatmaps, and download automated PDF reports.

Results show good overall performance with precision (~85%), recall (~83%), and F1-score (~82%). The model performs best on pituitary tumors but shows some confusion between glioma and meningioma due to similar imaging patterns. Compared to traditional machine learning and basic CNNs, NeuroScanAI performs better and offers added benefits of explainability and automation.

Conclusion

This paper presented \"NeuroScanAI\", a deep learning- based system for brain tumor detection using MRI images. The proposedsystem uses transfer learningwith CNN models to achieve an accuracy of approximately 83–84%. Grad-CAM visuals enhance model by highlighting tumor regions, while the web interface and automated PDF report generation improve practical use and make the system moreeffectivefor doctors, patients and real-world operations . The results demonstrate that the system provides an efficient and reliable solution for assisting medical personals.

References

[1] H.Kumar,N.Chavali,S.Patil,S.Ibrahim,and H. Sahoo, \"Automated Brain Tumor Identification and Categorization via Convolutional Neural Networks in Deep Learning,\"2024 Int, IEEE, pp. 433–437, 2024. [2] H. A. Mahzabin, A. Hoque, A. A. Azad, A. M. Rony, and F. N. Nipa, \"Transfer Learning and Explainable AI for Brain Tumor Classification: A Study Using MRI Data from Bangladesh,\"arXiv, 2024. Available: https://arxiv.org/abs/2405.12345 [3] N. Mahzabin, A. Chowdhury, and T. Das, \"Automatic Brain Tumor Classification Based on Transfer Learning Models,\" IEEE, pp. 136–140, 2022. [4] R. T. Arif, M. H. Mahmud, and S. Rahman, \"Comparative Analysis of Resource-Efficient CNN ArchitecturesforBrainTumorClassification,\"IEEE 2024.Available: https://arxiv.org/abs/2411.15596 [5] A. Sharma and R. Gupta, \"Deep CNN ArchitecturesforBrainTumorDetectionUsingMRI Scans,\"IEEE Access, vol. 9, pp. 115930–115940, 2021. [6] S. Kumar, R. Patel, and A. Singh, “Deep TransferLearningBasedBrainTumorClassification Using MRI Images,” IEEE2025. [7] M. H. Rahman, T. Islam, and A. A. Azad, “Explainable AI for Brain Tumor Detection Using Grad-CAM and CNN Models,” IEEE ,pp. 567–578, Feb 2025. [8] N. Das, K. Roy, and S. Banerjee, “MRI-Based Brain Tumor Classification Using EfficientNet and Transfer Learning,” IEEE 2026

Copyright

Copyright © 2026 Keshav Asawa, Ketan Gujare, Vraj Gohil, Keval Desai. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.