Machine learning approach for distinction of ADHD and OSA

The purpose of this study is to find an efficient way to discriminate between Attention-deficit/ hyperactivity disorder (ADHD) and Obstructive sleep apnea (OSA). The study collected 217 children (aged 6-12 years) data between 2011 and 2015, who were divided into three groups, ADHD, OSA and a combination of ADHD and OSA. Each group based on the doctor's determination, using the DSM-IV diagnostic standards. The data included four questionnaires as follow: CBCL, DBRS, OSA-18 and CSHQ. In order to speed up the whole process of clinical diagnosis classification, we train and test three machine learning models to find the best way to help clinical doctor to diagnosis. The study results indicate that in all of subscale items, there were 17 item show significantly difference among three subgroups, especially in the CBCL. Our results also show that CART model has better computational efficiency than CHAID and Neural Network model for subgroups classification.