Manuscript received July 25, 2025; revised August 14, 2025; accepted October 15, 2025; published April 28, 2026.

Abstract—Brain tumour detection and classification remain critical tasks in medical imaging. Although Convolutional Neural Networks (CNNs) have demonstrated significant potential, most studies rely on two-dimensional Magnetic Resonance Imaging (MRI) datasets, which may contain duplicated or redundant samples, leading to over-optimistic performance. In this paper, a new approach that utilises ensemble transfer learning is presented, integrating four well-known CNN architectures: EfficientNetB7, MobileNetV2, VGG16, and Xception. Each model is first fine-tuned through transfer learning, and their extracted feature maps are concatenated to form a richer representation before classification using dense layers and a softmax output layer. In addition to reporting the ensemble performance, thorough experiments are conducted to evaluate the contribution of each architecture individually and compare the proposed method against state-of-the-art CNNs and Vision Transformer (ViT)-based models. Results on the Kaggle7023 dataset demonstrate high classification accuracy, although some limitations related to the dataset are emphasised, particularly its 2D nature and potential redundancy, which restrict generalisation. Findings highlight both the promise and the challenges of ensemble deep learning models for uprediction, specifically predicting brain tumours.

Keywords—brain tumour, artificial intelligence, diagnosis, medical imaging

Cite: Istabraq H. Jasim, Zakariya A. Oraibi, and Entesar B. Talal, "Ensemble Transfer Learning Approach for Efficient Brain Tumour Prediction," Journal of Image and Graphics, Vol. 14, No. 2, pp. 303-311, 2026.

Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC-BY-4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.