基於機器學習分析ADHD亞型利用VR認知測驗同步的神經生理數據;Analysis of ADHD Subtypes Using Machine Learning with Synchronized Neurophysiological Data from VR Based Cognitive Tests

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/95718

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/95718

Title:	基於機器學習分析ADHD亞型利用VR認知測驗同步的神經生理數據;Analysis of ADHD Subtypes Using Machine Learning with Synchronized Neurophysiological Data from VR Based Cognitive Tests
Authors:	林廷真;Lin, Ting-Jhen
Contributors:	資訊工程學系
Keywords:	注意力缺陷/過動症(ADHD);亞型;認知測驗;持續表現測試(CPT);威斯康辛卡片分類測試(WCST);腦電圖(EEG);機器學習(ML);Attention Deficit/Hyperactivity Disorder (ADHD);subtypes;Cognitive tests;Continuous Performance Test (CPT);Wisconsin Card Sorting Test (WCST);Electroencephalography (EEG);Machine Learning (ML)
Date:	2024-08-01
Issue Date:	2024-10-09 17:11:37 (UTC+8)
Publisher:	國立中央大學
Abstract:	ADHD亞型是個值得關注的議題，現今的診斷大多主觀又耗時，為了實現精準醫療，根據個體化的數據來提供針對性的治療，本研究採用了在虛擬現實（VR）環境中進行的認知測驗，並同步收集測驗表現與神經生理數據：頭轉, 眼動和腦電，以統計分析和分組分類來探討ADHD亞型之間及亞型與正常組在各方面的差異。從結果得知，即便亞型與正常組間在外在表現上沒有太大的差異但能從腦電看出兩組在大腦活動的不同。而亞型間在單一測驗的頭轉眼動資料幾乎沒有顯著特徵，但在多測驗資料的分類卻能達86%的準確率。在分類上，只分有無ADHD的效果大多沒有比ADHD亞型間的分類或是控制組的比較來得好。另一個發現是，在腦電的頻域特徵分析中，在聽覺CPT的顯著特徵都是發生在點位Oz上，反映出非額葉上的腦電點位對檢測ADHD亞型也是有一定的重要性。在未來，期望這系統能為ADHD亞型提供客觀的數據，除了能成為輔助檢測ADHD亞型的工具之外，也能為未來精準醫療奠定基礎。;ADHD subtypes are a critical area of concern, with current diagnoses often being subjective and time-consuming. To achieve precision medicine and provide targeted treatments based on individualized data, this study utilized cognitive tests in a virtual reality (VR) environment. Tests performance data and neurophysiological data: head rotation, eye movement, and EEG, were synchronously collected and analyzed. Statistical analysis and classification were performed to explore differences between ADHD subtypes and between subtypes and the normal group. Results indicate that, despite minimal external behavioral differences between subtypes and the normal group, EEG data reveal distinct brain activity patterns. While individual tests showed few significant features in head rotation and eye movement, combining multiple tests achieved an 86% classification accuracy. Diagnosing ADHD as a whole was generally less effective than differentiating subtypes or comparing with controls. Additionally, frequency-domain analysis of EEG features from audio CPT highlighted significant findings at the Oz electrode site, underscoring the relevance of non-frontal EEG locations for detecting ADHD subtypes. In the future, this system is expected to provide objective data for ADHD subtypes, serving as a tool for subtype detection and laying the groundwork for precision medicine.
Appears in Collections:	[Graduate Institute of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	40	View/Open

社群 sharing

Loading...