Brain Tumor Classification from MRI Using Colormap-Based Vision Transformers on the BRISC2025 Dataset

Accurate brain tumor classification from magnetic resonance imaging (MRI) is essential for supporting early diagnosis and treatment planning. While Vision Transformers (ViTs) have recently demonstrated strong performance in medical image analysis through their ability to model long-range spatial dependencies, the influence of input feature representation on transformer-based brain tumor classification remains insufficiently explored. To address this gap, this study proposes ViT-Color, a colormap-enhanced Vision Transformer framework that transforms grayscale MRI scans into pseudo-color representations to emphasize subtle intensity variations and structural characteristics prior to classification. Unlike conventional ViT pipelines that directly utilize grayscale images, the proposed framework investigates whether colormap-based feature enhancement can improve the discriminative capability of transformer models for multi-class brain tumor recognition. The proposed approach was evaluated on the BRISC2025 dataset containing four diagnostic categories: glioma, meningioma, pituitary tumor, and non-tumorous brain MRI images. To ensure unbiased model selection, the official training set was partitioned into training and validation subsets, while the official test set remained completely unseen until final evaluation. Experimental results demonstrate that ViT-Color achieves an accuracy of 98.63, precision of 98.65, recall of 98.81, F1-score of 98.65, and an AUC of 99.95. Furthermore, an ablation study comparing grayscale and colormap-based inputs confirms that the proposed color-enhanced representation consistently improves classification performance across all evaluation metrics. These findings indicate that colormap-guided feature enhancement can effectively complement Vision Transformers for robust brain tumor classification and may provide a practical direction for improving transformer-based medical image analysis systems.