Copyright © 2026 Authors retain the copyright of this article. This article 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.

@article{200340,
        author = {Akula Vennela Priya and Dr. Ummadi Sathish Kumar and Sabbavarapu Siri Chandhana and Gompa Lokesh and Teppala Akhil},
        title = {Brain Tumor Detection From MRI Scans Using Deep-learning},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {12},
        pages = {1091-1096},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=200340},
        abstract = {Brain tumor diagnosis from Magnetic Resonance Imaging (MRI) is a critical yet time-intensive clinical task that demands highly accurate detection and precise spatial delineation of tumour boundaries. This paper presents NeuroScan AI, an end-to-end automated pipeline that combines classical image processing with deep learning to achieve robust brain tumor classification and segmentation across four categories: glioma, meningioma, pituitary adenoma, and healthy tissue. The classification stage employs a weighted ensemble of three convolutional neural networks — ResNet-50 with Convolutional Block Attention Module (CBAM), EfficientNet-B3, and DenseNet-121 — achieving an ensemble AUC of 0.9952. For segmentation, we depart from GradCAM-based pseudo-label generation, which is shown to produce anatomically incorrect activations, and instead propose a physically motivated classical pipeline based on brain extraction, CLAHE enhancement, multi-percentile intensity sweeping, and compactness-weighted blob scoring. These classical masks serve as pseudo ground-truth labels to supervise an Attention U-Net trained with a combined Tversky, Dice, and boundary-weighted loss. The segmentation pipeline achieves a validation Dice coefficient of 0.4181 on 240 images using a T4 GPU in under 10 minutes. Boundary extraction via cv2.fitEllipse produces smooth, clinically interpretable contours consistent with radiological annotation style. NeuroScan AI provides a practical, deployable framework for computer-aided brain tumor analysis.},
        keywords = {Brain Tumor, MRI, Deep Learning, Attention U-Net, Ensemble Learning, Image Processing, Segmentation.},
        month = {May},
        }