摘要: | 注意缺陷多動障礙 (ADHD) 是一種神經發育障礙,通常發生在兒童 時期。ADHD 的主要症狀是難以集中注意力、過動、性格衝動,甚至 有破壞性的行為。 ADHD 通常在學齡前被診斷出來。如果不及時有 效地治療,症狀可能會持續到青春期,甚至到成年期。因此,通過 高效、有效和低成本的認知評估來檢測多動症患者非常重要。很多 時候,ADHD 的評估完全依賴於使用 BRS 進行患者行為觀察和評分的 醫生和家長,因此這些評估方法容易具有很強的主觀性。由於上述 缺點,我們更喜歡使用虛擬現實(VR)技術。 VR 可以通過多種傳 感器提供身臨其境的交互虛擬環境,讓我們更好地獲取用戶信息。 在這項研究中,我們設計了一個虛擬課堂遊戲。遊戲內容基於音頻 測試、CPT 和 Stroop 測試,加上我們在遊戲中設計的一些干擾事 件,以便我們盡量獲取我們需要的用戶的生理信息。生理信息主要 包括遊戲中的任務表現、腦電波、眼球運動軌跡、頭部旋轉幅度 等,分別轉化為多維數據集。然後通過統計分析確定這些數據集的 特徵在不同類型的用戶之間是否存在顯著差異。此外,隨著機器學 習和深度學習的飛速發展,我們也利用上述技術幫助我們構建了一 個能夠區分正常人和多動症患者的分類模型。由於我們有不同的數 據集,我們可以使用每個數據集來構建模型並評估模型的有效性。 甚至可以組合所有數據集來構建融合模型。最終結果表明,我們的 實驗具有很大的發展潛力。;Attention deficit hyperactivity disorder (ADHD) is a neurodevelopmental disorder that usually occurs in childhood. The main symptoms of ADHD are inattention, hyperactivity, impulsivity, and even destructive behaviors. ADHD is usually diagnosed before school age. If it is not treated effectively and timely, the problems will continue into adolescence and even into adulthood. Therefore, it is very important to screen people with ADHD through efficient, effective and low-cost cognitive assessment. In many times, the assessment of ADHD completely relies on doctors and parents who use BRS for patient behavior observation and rating, so these assessment methods prone to have a strong subjectivity. Due to the above shortcomings, we prefer to use Virtual Reality (VR) technology. VR can provide an immersive and interactive virtual environment with a variety of sensors so that we can better obtain user information. In this research, we designed a virtual classroom game. The content of the game is based on audio test, CPT and Stroop test with some interference events we designed in the game, so that we can try our best to get the physiological information of the user we need. The physiological information mainly includes the task performance in the game, brain waves, eye movement trajectory, and head rotation amplitude, which are respectively converted into a multi-dimensional dataset. Then determine whether the features of these datasets are significantly different among different types of users through statistical analysis. In addition, with the rapid development of machine learning and deep learning, we have also used the above technologies to help us build a classification model that can distinguish between normal people and ADHD patients. Since we have diverse datasets, we can use each dataset to build a model and evaluate the effectiveness of the model. It is even possible to combine all the datasets to build a fusion model. The final results show that our experiment has great potential for development. |