Scholarlink

<p>This study addresses the problem of automatic binary classification of experimental participants diagnosed with schizophrenia and a control group using a dataset obtained using a Siemens Magnetom Verio 3T tomograph. The dataset included data from 36 experimental participants undergoing treatment at Clinical Hospital No. 1, Moscow Health Department, and 36 experimental participants from the control group. Machine learning methods were used to address the study task. A separation accuracy of 76% was achieved, which is comparable with results obtained in other scientific studies. The highest accuracy was obtained for the local homogeneity parameter (<i>ReHo</i>), which is already known in the literature. The set of features developed by the authors based on the method of identifying functionally homogeneous regions (<i>FHR</i>) produced a maximum classification accuracy of 74%. However, the set of <i>FHR</i> features provides higher classification accuracy when a small number of brain regions is used. For example, use of just eight regions led to the <i>FHR</i> set providing near-maximal classification accuracy, 72.5% (versus 65% for the <i>ReHo</i> set), which suggests that these selected eight regions provide the highest level of separation.</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

The Search for the Most Informative Areas for the Binary Classification of Schizophrenia Using Resting fMRI Data Based in a Method for Extracting Functionally Homogeneous Areas

  • 期刊论文

  • 发表日期：

    2025年1月1日

查看全文
  • A. D. Zhemchuzhnikov,
  • S. I. Kartashov,
  • S. O. Kozlov,
  • V. A. Orlov,
  • A. A. Poyda,
  • N. V. Zakharova,
  • L. V. Bravve,
  • G. Sh. Mamedova,
  • M. A. Kaydan

摘要

This study addresses the problem of automatic binary classification of experimental participants diagnosed with schizophrenia and a control group using a dataset obtained using a Siemens Magnetom Verio 3T tomograph. The dataset included data from 36 experimental participants undergoing treatment at Clinical Hospital No. 1, Moscow Health Department, and 36 experimental participants from the control group. Machine learning methods were used to address the study task. A separation accuracy of 76% was achieved, which is comparable with results obtained in other scientific studies. The highest accuracy was obtained for the local homogeneity parameter (ReHo), which is already known in the literature. The set of features developed by the authors based on the method of identifying functionally homogeneous regions (FHR) produced a maximum classification accuracy of 74%. However, the set of FHR features provides higher classification accuracy when a small number of brain regions is used. For example, use of just eight regions led to the FHR set providing near-maximal classification accuracy, 72.5% (versus 65% for the ReHo set), which suggests that these selected eight regions provide the highest level of separation.