Scholarlink

Alzheimer disease is the major cause for dementia arising due to the cognitive decline and neurodegeneration. Early prediction is important for better treatment and the complexity of disease requires multiple modalities such as MRI, PET, and clinical data. Conventional approaches used for diagnosis utilize these modalities independently limiting the diagnostic capability. Considering this limitation, a transformer-based multimodal fusion method utilizing cross-attention mechanism has been proposed to enhance diagnostic accuracy. By utilizing the cross-attention mechanism, the model has ability to identify important features from each modality which helps in fusing complementary information for better diagnosis. This approach not only enhances predictive accuracy but also provides valuable insights into the mechanisms of Alzheimer’s disease prediction. Utilization of dropout regularization method helps the model to converge faster making the proposed method highly effective. Combining the strengths of multimodal fusion and cross-attention, the proposed method can be utilized for early Alzheimer’s disease diagnosis, leading to better patient outcomes and accuracy in Alzheimer’s disease diagnosis.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multimodal Alzheimer’s Disease Diagnosis Using Cross-Attention Transformers and Dense Networks with Dropout Regularization

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • J. Aswini,
  • Konduru Venkateswara Raju,
  • V. Lavanya,
  • Chekuri Nalini,
  • Sreekanth Yalavarthi,
  • Naresh Tangudu

摘要

Alzheimer disease is the major cause for dementia arising due to the cognitive decline and neurodegeneration. Early prediction is important for better treatment and the complexity of disease requires multiple modalities such as MRI, PET, and clinical data. Conventional approaches used for diagnosis utilize these modalities independently limiting the diagnostic capability. Considering this limitation, a transformer-based multimodal fusion method utilizing cross-attention mechanism has been proposed to enhance diagnostic accuracy. By utilizing the cross-attention mechanism, the model has ability to identify important features from each modality which helps in fusing complementary information for better diagnosis. This approach not only enhances predictive accuracy but also provides valuable insights into the mechanisms of Alzheimer’s disease prediction. Utilization of dropout regularization method helps the model to converge faster making the proposed method highly effective. Combining the strengths of multimodal fusion and cross-attention, the proposed method can be utilized for early Alzheimer’s disease diagnosis, leading to better patient outcomes and accuracy in Alzheimer’s disease diagnosis.