Colormap-Enhanced Vision Transformers for MRI-Based Multiclass (4-Class) Alzheimer's Disease Classification

TL;DR

The paper introduces PseudoColorViT-Alz, a colormap-enhanced Vision Transformer for four-class MRI-based Alzheimer's classification. By converting grayscale MRIs to pseudo-color using a jet colormap, the method leverages ViT’s global self-attention to better capture subtle structural cues, achieving state-of-the-art performance on OASIS-1 with 99.79% accuracy and a perfect AUC of 100%. The approach demonstrates that pseudo-color representations can significantly boost discriminative power in medical imaging when paired with pretrained transformers, offering a robust tool for early detection and clinical decision support. Limitations include validation on additional datasets and exploration of other colormaps to further improve generalization and interpretability.

Abstract

Magnetic Resonance Imaging (MRI) plays a pivotal role in the early diagnosis and monitoring of Alzheimer's disease (AD). However, the subtle structural variations in brain MRI scans often pose challenges for conventional deep learning models to extract discriminative features effectively. In this work, we propose PseudoColorViT-Alz, a colormap-enhanced Vision Transformer framework designed to leverage pseudo-color representations of MRI images for improved Alzheimer's disease classification. By combining colormap transformations with the global feature learning capabilities of Vision Transformers, our method amplifies anatomical texture and contrast cues that are otherwise subdued in standard grayscale MRI scans. We evaluate PseudoColorViT-Alz on the OASIS-1 dataset using a four-class classification setup (non-demented, moderate dementia, mild dementia, and very mild dementia). Our model achieves a state-of-the-art accuracy of 99.79% with an AUC of 100%, surpassing the performance of recent 2024--2025 methods, including CNN-based and Siamese-network approaches, which reported accuracies ranging from 96.1% to 99.68%. These results demonstrate that pseudo-color augmentation combined with Vision Transformers can significantly enhance MRI-based Alzheimer's disease classification. PseudoColorViT-Alz offers a robust and interpretable framework that outperforms current methods, providing a promising tool to support clinical decision-making and early detection of Alzheimer's disease.

Figures (4)

• Figure 1: Representative brain MRI samples from the OASIS-1 dataset illustrating the four Alzheimer’s disease categories used in this study.
• Figure 2: PseudoColorViT-Alz preprocessing pipeline. Overview of the complete data preprocessing workflow, illustrating the steps from raw brain MRI image acquisition, grayscale conversion, and pseudo-color enhancement to normalization, dataset construction, and input preparation for Vision Transformer–based multiclass Alzheimer’s disease classification.
• Figure 3: PseudoColorViT pipeline. The proposed framework follows the Vision Transformer architecture introduced in dosovitskiy2021image. A 2D grayscale medical image is first converted into a pseudo-color representation, yielding a three-channel input that preserves anatomical structure while enhancing texture and contrast information. The pseudo-colored image is then partitioned into fixed-size patches, which are linearly projected and augmented with positional encodings. A learnable classification token is appended to the patch sequence and processed by the Transformer encoder. The final encoded representation is passed through a classification head to produce the disease prediction.
• Figure 4: Performance evaluation of PseudoColorViT-Alz on the OASIS-1 dataset. The left panel shows one-vs-rest ROC curves and corresponding AUC values for each Alzheimer’s disease class, while the right panel presents the confusion matrix illustrating class-wise prediction performance.