| Summary: | 碩士 === 國立陽明大學 === 生物醫學資訊研究所 === 103 === Objective: This study used a machine learning approach to identify the common or unique features from the attention performance to distinguish children with attention deficit/hyperactivity disorder (ADHD) from those without, and to determine which items would improve or decrease the prediction accuracy of ADHD.
Method: This study included 799 children with ADHD, aged 7-18 years old, and 421 same-aged controls. Their attention performance assessed by the Conners’ Continuous Performance Test (CCPT), and ADHD-related symptoms measured by the Chinese Version of the Swanson, Nolan, and Pelham IV Scale (SNAP-IV)-Parent and Teacher Forms, and the Conners’ Parent and Teacher Rating Scale-revised Short Form (CPRS and CTRS) were collected. Support vector machine was then used for data analysis.
Result: Through combinations of the features from these scales, we identified 9 features that can increase accuracy and 9 features can decrease accuracy when they had or had not been selected into machine learning.
Conclusion: The neuropsychological and self-administered measures predicting ADHD diagnosis may be improved after the approaches by machine learning. In this study we found features to improve previous scales or CCPT to diagnose ADHD children with high accuracy.
|
|---|
