Classification of Alzheimer’s disease Stages using Vision Tranformers

Abstract

Alzheimer’s disease progresses as a neurodegenerative disorder that creates substantial obstacles during early diagnosis and therapeutic intervention. Brain Magnetic ResonanceImaging(MRI)offerspromising diagnostichelp but requiresadvancedanalyticalmethodstodetectthedisease\'s subtle structural changes. New deep learning models which include Vision Transformers (ViTs) now receive significant attention due to their superior performance in medical imaging applications. Vision Transformersapply self-attention mechanisms to capture long-range data relationships which enhances their performance in brain MRI scan analysis. Researchers investigated how the ViT base architecture can differentiate between Alzheimer\'s disease stages.

Introduction

The World Health Organization (WHO) reports that dementia affects 55 million people worldwide, with Alzheimer’s disease (AD) accounting for 60–70% of cases. AD is a degenerative brain disorder mostly affecting those over 65, characterized by memory loss and cognitive decline. Its exact cause is unknown but likely involves genetic and environmental factors. Although no cure exists, treatments can alleviate symptoms and improve quality of life. AD progresses slowly, starting with a preclinical phase that can last 20 years, moving through Mild Cognitive Impairment (MCI) to dementia. Early diagnosis is crucial for better management.

AD detection methods include invasive biological marker testing and non-invasive medical imaging such as MRI, with MRI being superior for brain tissue visualization. Recent advances employ deep learning models, especially Vision Transformers (ViTs), which outperform traditional convolutional neural networks (CNNs) by capturing spatial relationships effectively. However, ViTs require large datasets or pre-trained models to function optimally.

The research focus is on classifying Alzheimer’s disease stages using brain MRI scans and Vision Transformers. The study reviews prior work using ViTs and hybrid models combining ViTs with other architectures (e.g., Bi-LSTM, CNN) for AD diagnosis and progression prediction. The methodology involves fine-tuning a pre-trained ViT model (vit-base-patch16-224) on MRI data from the ADNI database, emphasizing the hippocampus region, which is closely linked to AD.

Conclusion

Theauthorsintroduceaclassificationsystemdesignedto identify the stages of Alzheimer’s disease by analyzing brain MRI images using Vision Transformers (ViTs). This research applies ViTs to leverage their capabilities for detecting these stages with notable accuracy.The approach demonstrated strong performance, achieving 88% accuracy and an AUC of 0.92 on the Axial dataset. To enhance interpretability, we also generated detailed PDF reports corresponding to each predicted class. These findings highlight the diagnostic potential of VisionTransformersinidentifyingAlzheimer’sdisease. Supportedbydeeplearningoptimizationtechniquesand attentionmaps,ourmethodsimplifiestheunderstanding of key brain regions involved in the disease. By integrating interpretive reports with classification outputs,thesystemoffersvaluableinsights.Overall,this researchpresentsaninnovativeapproachtoAlzheimer’s diagnosis,showingpromiseasacomputer-aidedtoolfor improvedclinicaldecisionsandbetterpatientoutcomes.

References

[1] Vision transformers for the prediction of mild cognitive impairment to Alzheimer’s disease progression using mid-sagittal sMRI,Hoang, Gia Minh and Kim, Ue-Hwan and Kim, Jae Gwan,2023 [2] Transfer learning assisted classification and detection of Alzheimer’sdiseasestagesusing3DMRIscans,Maqsood,Muazzam andNazir,FariaandKhan,UmairandAadil,FarhanandJamal,HabibullahandMehmood,IrfanandSong,Oh-young,2019 [3] Efficiently Training Vision Transformers on Structural MRI Scans for Alzheimer’s Disease Detection, Dhinagar, Nikhil J and Thomopoulos, Sophia I and Laltoo, Emily and Thompson, Paul M,2023 [4] Addformer: Alzheimer’s disease detection from structural mri using fusion transformer ,Kushol, Rafsanjany and Masoumzadeh, Abbas and Huo, Dong and Kalra, Sanjay and Yang, Yee-Hong ,2023 [5] OViTAD:Optimizedvision transformertopredictvarious stagesofAlzheimer’sdisease usingresting-state fMRI and structuralMRI data, Sarraf, Saman and Sarraf, Arman and DeSouza, Danielle D and Anderson, John AE and Kabia, Milton and Alzheimer’s Disease Neuroimaging Initiative, 2023 [6] Vision Transformers and Bi-LSTM for Alzheimer’s Disease Diagnosisfrom3DMRI,Akan,TaymazandAlp,SaitandBhuiyanb,MohammadAN,2024 [7] Diagnosis of Alzheimer’s disease via optimized lightweight convolution-attentionandstructuralMRI,Khatri,UttamandKwon,Goo-Rak,2024 [8] DeeplearningforAlzheimer’sdiseasediagnosis:Asurvey,2022 [9] Gradient-basedlearningappliedtodocumentrecognition,1998 [10] Redmon, Joseph and Divvala, Santosh and Girshick, Ross and Farhadi, Ali, You Only Look Once: Unified, Real-Time Object Detection,2016 [11] Deeplearningbased pipelinesforAlzheimer’sdisease diagnosis:Acomparativestudyandanoveldeep-ensemblemethod,2022 [12] Silva,IagoR.R.andSilva,GabrielaS.L.anddeSouza,RodrigoG. anddosSantos,WellingtonP.anddeA.Fagundes,RobertaA,Model Based on Deep Feature Extraction for Diagnosis of Alzheimer’s Disease,2019 [13] Zhao, BendongandLu, HuanzhangandChen, Shangfeng andLiu,Junliang and Wu, Dongya, Convolutional neural networks for time seriesclassification,2017 [14] Vaswani, Ashish and Shazeer, Noam and Parmar, Niki and Uszkoreit,JakobandJones,LlionandGomez,AidanNandKaiser,?ukaszandPolosukhin,Illia,AttentionisAllyouNeed,2017 [15] Advit: Vision transformer on multi-modality pet images for alzheimerdiseasediagnosis,Xing,XinandLiang,GongboandZhang, Yu and Khanal, Subash and Lin, Ai-Ling and Jacobs, Nathan, 2022 [16] https://ida.loni.usc.edu/login.jsp?project=ADNI [17] Dosovitskiy,A.,Beyer,L.,Kolesnikov,A.,Weissenborn,D.,Zhai, X.,Unterthiner,T.,Dehghani,M.,Minderer,M.,Heigold,G.,Gelly,S., Uszkoreit, J., Houlsby, N. (2020).

Copyright

Copyright © 2025 Bakshish Singh, Dhanraj Singh, Aditya Vikram Singh, Tushar Singh. 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.