Authors:
Qaiser Sajawal, Rejwan Bin Sulaiman

Addresses:
Department of Computer Science and Technology, University of West Scotland, Paisley, Scotland, United Kingdom.

Abstract:

Despite the rapid progress in artificial intelligence and medical image analysis, the early diagnosis of brain tumors remains a major clinical challenge. Accurate tumor detection from magnetic resonance imaging (MRI) and computed tomography (CT) scans is critical for timely clinical intervention and better patient outcomes. In this work, researchers propose an interpretable cross-platform deep learning framework for automated brain tumor detection and segmentation, deployable on mobile and desktop devices. The proposed pipeline is a fine-tuned CNN (VGG16 architecture) for multi-class tumor classification followed by a U-Net architecture for precise tumor segmentation. Transfer learning is used to improve network performance for MRI and CT imaging modalities. The entire system is developed using the Flutter framework, which enables easy deployment across Android, iOS, Windows, macOS and the web, with real-time inference capabilities. To improve clinical transparency and user confidence, researchers generate visual explanations that identify diagnostically relevant regions of the input that impact the model prediction using Gradient-weighted Class Activation Mapping (Grad-CAM). The experimental evaluation shows better performance, with classification accuracies of 98.3% and 96.8% and Dice similarity coefficients of 0.938 and 0.921 for MRI and CT images, respectively, in tumour segmentation. This framework provides a great bridge between state-of-the-art deep learning research and real-world clinical deployment, combining excellent diagnostic accuracy, explainable artificial intelligence, efficient segmentation, and easy cross-platform accessibility for neuroradiology applications.

Keywords: Artificial Intelligence (AI); Brain Tumor Detection; Medical Image Analysis; Tumor Segmentation; Computed Tomography (CT); Cross-Platform Deployment; Transfer Learning.

Received on: 25/06/2025, Revised on: 28/08/2025, Accepted on: 07/11/2025, Published on: 18/08/2026

DOI: 10.69888/FTSHSL.2026.000730

FMDB Transactions on Sustainable Health Science Letters, 2026 Vol. 4 No. 3, Pages: 167-201

• Views : 35
• Downloads : 8

Download PDF