Scholarlink

The prompt identification of Autism Spectrum Disorder (ASD), a multifaceted neurological and developmental disease, presents a considerable barrier. This study investigates the detection of Autism Spectrum Disorder (ASD) by utilising functional connectivity analysis of resting-state functional MRI (rs-fMRI) data, employing machine learning approaches to use the functional brain network as a valuable information source. Hierarchical Support Vector Machine (H-SVM) and Interactive Embedding-K Nearest Neighbours (InEm-KNN) classifiers are utilised to differentiate between ASD patients and normally developing (TD) participants in the ABIDE dataset. The Support Vector Machine classifier utilising a radial basis function kernel attains the best accuracy of 67.97%, above that of alternative kernel functions. Comparisons between SVM and KNN models indicate that the meticulously calibrated Hierarchical SVM (H-SVM) model surpasses other leading approaches. A stacked ensemble model, integrating logistic regression, meticulously optimised Hierarchical SVM, and precisely calibrated Interactive Embedding KNN, is examined. Although the stacked model surpasses the independently optimised InEm-KNN, the independently optimised Hierarchical SVM model exhibits greater performance. The study reveals that a solo, meticulously optimised H-SVM classifier outperforms neural network topologies, attaining an accuracy of up to 89%.

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

Unraveling Autism Through FCN Using Hierarchical Support Vector Machine (H-SVM) and Interactive Embedding-K Nearest Neighbors (InEm-KNN)

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Srinithi Mitra,
  • Utkarsh Patel,
  • Sricheta Parui

摘要

The prompt identification of Autism Spectrum Disorder (ASD), a multifaceted neurological and developmental disease, presents a considerable barrier. This study investigates the detection of Autism Spectrum Disorder (ASD) by utilising functional connectivity analysis of resting-state functional MRI (rs-fMRI) data, employing machine learning approaches to use the functional brain network as a valuable information source. Hierarchical Support Vector Machine (H-SVM) and Interactive Embedding-K Nearest Neighbours (InEm-KNN) classifiers are utilised to differentiate between ASD patients and normally developing (TD) participants in the ABIDE dataset. The Support Vector Machine classifier utilising a radial basis function kernel attains the best accuracy of 67.97%, above that of alternative kernel functions. Comparisons between SVM and KNN models indicate that the meticulously calibrated Hierarchical SVM (H-SVM) model surpasses other leading approaches. A stacked ensemble model, integrating logistic regression, meticulously optimised Hierarchical SVM, and precisely calibrated Interactive Embedding KNN, is examined. Although the stacked model surpasses the independently optimised InEm-KNN, the independently optimised Hierarchical SVM model exhibits greater performance. The study reveals that a solo, meticulously optimised H-SVM classifier outperforms neural network topologies, attaining an accuracy of up to 89%.