International Journal of Computer Theory and Engineering

Abstract—Medical imaging is essential for diagnosing neurological diseases. Advances in Deep Learning (DL) have greatly improved brain Magnetic Resonance Imaging (MRI) classification, enabling more accurate anomaly detection. However, the high computational and memory demands of DL models pose challenges for deployment on resource-constrained platforms such as portable medical devices and edge computing systems. This study aims to address these limitations by reducing storage and transmission demands through deep learning-based tumour classification on compressed MRI data. JPEG2000 lossless compression was applied at varying ratios to examine its effect on classification performance. The analysis focuses on understanding how different compression levels impact the accuracy and reliability of DL models. Three deep learning architectures—Convolutional Neural Network (CNN), ResNet50, and MobileNetV2—were selected to represent baseline, high-capacity, and lightweight models, respectively, and were trained on compressed MRI datasets to classify brain images into glioma, meningioma, pituitary tumor, and no tumor. Evaluation metrics included Peak Signal-to-Noise Ratio (PSNR), entropy, and Structural Similarity Index (SSIM) for image quality, and precision, recall, accuracy, F1 score, and Area Under the Receiver Operating Characteristic Curve (AUROC) for classification performance. The experimental results indicate that CNN and ResNet50 exhibit higher classification metrics at specific compression levels. However, MobileNetV2 consistently maintains acceptable performance across all tested compression ratios. MobileNetV2 features a lightweight architecture with minimal memory requirements, making it well-suited for deployment in point-of-care devices. The model was optimized and deployed on the NVIDIA Jetson TX2 Developer Board, enabling the development of a portable, lightweight diagnostic tool for brain tumor detection.

Keywords—medical imaging, MRI brain classification, deep learning optimization, model compression, edge Artificial Intelligence (AI) deployment, lightweight neural networks

[PDF]

Cite: Neena K A and Anil Kumar M N, "Optimizing Medical MRI Brain Image Classification through Compression Analysis on Deep learning Models with Light Weight Implementation," International Journal of Computer Theory and Engineering, vol. 18, no. 1, pp. 68-78, 2026.

Copyright © 2026 by the authors. 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 (CC BY 4.0).