Scholarlink

Cracks and crystal-like formations observed in desiccated biological fluids present diagnostic opportunities for a variety of diseases. This work aims to study the presence of diagnostic biomarkers in relation with early preclinical stages of Alzheimer’s disease, in dried human samples from cerebrospinal fluid and plasma. Employing radiomic features extracted from microscope images of dried droplets in both fluids, we build machine learning and deep learning classifiers to predict the presence of the disease. Gradient Boosting and Extra Tree classifiers perform best, yielding an area under the curve of 77% in cerebrospinal fluid, 76% in plasma, and 83% when combined. These results corroborate the potential of this technique as a complementary approach in the development of cost-effective diagnostic tools for Alzheimer’s disease when using non-invasive biofluids such as plasma.

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

Alzheimer’s Disease Classification by Artificial Intelligence Using Microscopy Images of Dried Human Body Fluids

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Javier Bóbeda,
  • Jaione Etxebarria-Elezgarai,
  • Laura Valeria Perez-Herrera,
  • María Jesús García-González,
  • Eduardo Alonso,
  • Maite García-Sebastián,
  • Miren Altuna,
  • Mirian Ecay-Torres,
  • Ainara Estanga,
  • Mikel Tainta,
  • Carolina López,
  • Eneko Lopez,
  • Pablo Martínez-Lage,
  • Andreas Seifert,
  • Karen López-Linares

摘要

Cracks and crystal-like formations observed in desiccated biological fluids present diagnostic opportunities for a variety of diseases. This work aims to study the presence of diagnostic biomarkers in relation with early preclinical stages of Alzheimer’s disease, in dried human samples from cerebrospinal fluid and plasma. Employing radiomic features extracted from microscope images of dried droplets in both fluids, we build machine learning and deep learning classifiers to predict the presence of the disease. Gradient Boosting and Extra Tree classifiers perform best, yielding an area under the curve of 77% in cerebrospinal fluid, 76% in plasma, and 83% when combined. These results corroborate the potential of this technique as a complementary approach in the development of cost-effective diagnostic tools for Alzheimer’s disease when using non-invasive biofluids such as plasma.