Scholarlink

<p>Alzheimer’s dementia is a progressive and irreversible neurological disorder that is a common cause of cognitive decline in humans, affecting more than 50 million people worldwide. Early detection of AD is critical as it helps slow the progression of this disease and improves the overall quality of life of an Alzheimer’s patient. In this paper, we propose <b>MEFAAD-U</b>, a novel machine learning-based <b>M</b>ultilingual <b>E</b>arly <b>F</b>usion <b>A</b>pproach for <b>A</b>lzheimer’s <b>D</b>ementia (AD) detection framework, with an emphasis on <b>U</b>rdu. We have leveraged non-invasive speech analysis due to the lack of cost-effective traditional diagnostic methods in this domain. Unlike previous English-centric models, we have collected and pre-processed an Urdu speech dataset as part of a multilingual corpus that also includes English, Greek, Spanish, and German. Urdu as a linguistically underrepresented language in AD detection research, is a key focus of this study. A distinguishing aspect of our framework is the early fusion of acoustic and linguistic features, which enhances the discriminatory power of the model. We evaluated MEFAAD-U using machine learning models such as support vector machines, linear regression, etc. The most promising results were obtained through the early fusion of features, followed by dimensionality reduction using Principal Component Analysis (PCA). Our best-performing model—an SVM trained on the fused feature set—achieved an accuracy of 0.85. This research highlights the potential of speech-based cross-linguistic tools for early diagnosis of AD. It is an advancement in medical diagnostics that leads to a significant social impact, providing global healthcare accessibility to underprivileged populations.</p>

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

Multilingual early fusion approach for Alzheimer’s dementia detection, including Urdu

  • 期刊论文

  • 发表日期：

    2026年2月27日

查看全文
  • Aniqa Hanif,
  • Khadija Saleem,
  • Ahmad Ullah Khan,
  • Aamir Wali,
  • Amina Javaid

摘要

Alzheimer’s dementia is a progressive and irreversible neurological disorder that is a common cause of cognitive decline in humans, affecting more than 50 million people worldwide. Early detection of AD is critical as it helps slow the progression of this disease and improves the overall quality of life of an Alzheimer’s patient. In this paper, we propose MEFAAD-U, a novel machine learning-based Multilingual Early Fusion Approach for Alzheimer’s Dementia (AD) detection framework, with an emphasis on Urdu. We have leveraged non-invasive speech analysis due to the lack of cost-effective traditional diagnostic methods in this domain. Unlike previous English-centric models, we have collected and pre-processed an Urdu speech dataset as part of a multilingual corpus that also includes English, Greek, Spanish, and German. Urdu as a linguistically underrepresented language in AD detection research, is a key focus of this study. A distinguishing aspect of our framework is the early fusion of acoustic and linguistic features, which enhances the discriminatory power of the model. We evaluated MEFAAD-U using machine learning models such as support vector machines, linear regression, etc. The most promising results were obtained through the early fusion of features, followed by dimensionality reduction using Principal Component Analysis (PCA). Our best-performing model—an SVM trained on the fused feature set—achieved an accuracy of 0.85. This research highlights the potential of speech-based cross-linguistic tools for early diagnosis of AD. It is an advancement in medical diagnostics that leads to a significant social impact, providing global healthcare accessibility to underprivileged populations.