Citation

Haq, Ikram Ul and Ullah, Asad and Khatoon, Amna and Khan, Adil and Ahmad, Shabeer and Roslee, Mardeni (2025) Transferable and Explainable Deep Learning Framework for Brain Tumor and COVID-19 Pneumonia Detection from Medical Imaging. In: 2025 Multimedia University Engineering Conference (MECON), 21-23 July 2025, Cyberjaya, Malaysia.

Text 485.pdf - Published Version Restricted to Repository staff only Download (1MB)

Abstract

Timely and reliable interpretation of radiological imaging is critical for effective clinical decision-making; however, manual analysis is labor-intensive, time-consuming, and susceptible to significant inter-observer variability. Advances in deep learning, particularly convolutional neural networks (CNNs), offer promising solutions to automate and improve diagnostic accuracy. This study introduces a transferable deep learning framework specifically designed to perform two distinct and clinically relevant medical image classification tasks: the detection of intracranial tumors in axial T1-weighted magnetic resonance imaging (MRI) scans and the identification of COVID-19 pneumonia from posterior-anterior chest radiographs (X-rays). The proposed approach employs a ResNet-50 architecture pre-trained on ImageNet, leveraging transfer learning to address data scarcity issues commonly encountered in medical imaging. The model used a lot of data augmentation and class-balanced sampling. A well-planned staged fine-tuning process was also used. These methods help stop overfitting and make the model better at working with new data. Model interpretability was added through Grad-Cam, a method that shows which parts of the images the model used most for its predictions. This technique provides a clear visual of key regions that influenced the diagnostic results. The framework performed very well on separate test data. It reached an accuracy of 99.7% for finding brain tumors. For COVID-19 pneumonia, it achieved an accuracy of 99.9%. Saliency maps help to clearly show how the model makes its decisions, making them easier to understand. This study highlights how combining transfer learning with explainable AI can improve diagnosis. These methods can make healthcare more accurate and understandable, especially where resources are limited.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Deep Learning, Medical Image Classification, Transfer Learning, Explainable AI (XAI), Brain Tumor and COVID-19 Detection
Subjects:	R Medicine > RC Internal medicine > RC0254 Neoplasms. Tumors. Oncology (including Cancer)
Divisions:	Faculty of Engineering (FOE)
Depositing User:	Ms Suzilawati Abu Samah
Date Deposited:	17 Mar 2026 03:35
Last Modified:	19 Mar 2026 01:39
URII:	http://shdl.mmu.edu.my/id/eprint/15475

Downloads

Edit (login required)