| Attention-Deficit Hyperactivity Disorder (ADHD) is the most prevalent neuropsychiatric disorder in children and adolescents, affecting 3–4% of school children in the United States [89]. Saccades are rapid eye movements that bring new and interesting visual targets onto the fovea of the retina for closer inspection. The regions of the brain responsible for controlling saccades are well understood and overlap those regions implicated in ADHD. It has been hypothesized that this overlap in brain areas will result in altered saccade performance in ADHD subjects. This thesis attempts to apply an appropriate case-based reasoning paradigm, based on the TA3 system, to the ADHD saccade performance data collected by Dr. Munoz and colleagues [66]. The primary goal of this study is to develop a reproducible, systematic approach to applying case-based reasoning to a new classification problem and apply this to the saccade performance data. It was hypothesized that there is sufficient information contained in the ADHD saccade performance data for the TA3 system to perform diagnosis with high accuracy. An iterative refinement methodology was developed, based on previous research in the field, and it was shown that this methodology was successful in developing successive CBR models with increasing performance, in a weak-theory domain. It was also shown that saccade performance variables could be used to distinguish ADHD from control children with an accuracy of over 70%. The relative importance of these variables was determined and this is a step towards using such metrics in a clinical setting. |
