Scholarlink

Alzheimer’s Disease (AD) is a neurodegenerative disorder that profoundly affects the hippocampus, a brain region critical for memory and cognition. This study leverages advanced imaging and radiomics techniques to identify biomarkers for AD by analyzing 3D MRI data. The methodology comprises two main steps: (1) manual segmentation of the hippocampus using 3D Slicer software, focusing on the axial plane where the region is most distinct, and (2) extraction of 104 radiomics features, including first-order, shape, texture, and wavelet-based features, using the Pyradiomics library. To determine the most effective subset of features, combinations of feature ranges (referred to as views) were systematically evaluated and compared against individual feature groups. Three approaches were employed: treating all features as a single dataset with random splitting (MVL Random Split), evaluating all views collectively (MVL View-Wise) and selecting the best performing feature sets(Best Selected view). Performance was assessed using metrics and the results reveal that the Best Selected View outperformed all other approaches, achieving superior performance with an accuracy of 0.78, AUC of 0.83, F1 score of 0.76, specificity of 0.86, and sensitivity of 0.73. This approach demonstrated significant improvements over the View-Wise method and the Random Split approach, as well as individual feature groups such as GLDM, Shape_2D, and first-order features. The findings highlight the critical role of feature selection and the effectiveness of combining multiple views to enhance predictive performance.

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

Optimal Radiomics Feature Selection to Enhance the Alzheimer’s Disease Detection Using Multi-view Learning Approach Based-on Hippocampal MRI Data

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ibrahim Momin,
  • Vipin Kumar,
  • Akshansh Gupta,
  • Bharti Rana

摘要

Alzheimer’s Disease (AD) is a neurodegenerative disorder that profoundly affects the hippocampus, a brain region critical for memory and cognition. This study leverages advanced imaging and radiomics techniques to identify biomarkers for AD by analyzing 3D MRI data. The methodology comprises two main steps: (1) manual segmentation of the hippocampus using 3D Slicer software, focusing on the axial plane where the region is most distinct, and (2) extraction of 104 radiomics features, including first-order, shape, texture, and wavelet-based features, using the Pyradiomics library. To determine the most effective subset of features, combinations of feature ranges (referred to as views) were systematically evaluated and compared against individual feature groups. Three approaches were employed: treating all features as a single dataset with random splitting (MVL Random Split), evaluating all views collectively (MVL View-Wise) and selecting the best performing feature sets(Best Selected view). Performance was assessed using metrics and the results reveal that the Best Selected View outperformed all other approaches, achieving superior performance with an accuracy of 0.78, AUC of 0.83, F1 score of 0.76, specificity of 0.86, and sensitivity of 0.73. This approach demonstrated significant improvements over the View-Wise method and the Random Split approach, as well as individual feature groups such as GLDM, Shape_2D, and first-order features. The findings highlight the critical role of feature selection and the effectiveness of combining multiple views to enhance predictive performance.