Early Detection of ADHD in Children Using Two Learning Approaches

Children with Attention Deficit /Hyperactivity Disorder (ADHD) struggle to focus for longer than a few minutes at a time. The onset of symptoms can occur as early as age 3 or 12, and they may worsen until maturity. For symptoms to be minimised and behavioural techniques to be improved, early diagnosis is critical. Even though it was first identified in the 1940s, after the 1990s, awareness grew. ADHD detection may also be impacted by genetic issues. Boys are treated more often than girls, which makes them less attentive and makes it harder for them to develop objectives. Brain analyses and behavior-related questionnaires are two early diagnostic techniques. The proposed system composed of two phases of diagnosing the ADHD in children. The first phase uses the dataset obtained from screening test in which Random Forest gave the highest accuracy. The second phase uses the Deep Neural Network with attention (DNN-AM) model for the early identification of ADHD in children. It analyses images as a whole, tracking how children with ADHD interact with their surroundings. It employs many modules for unique patterns and a single DNN-AM network for feature extraction and classification. By employing a Local Binary Pattern image for analysis, the feature set is expanded, improving the extraction of ADHD patterns in children. The suggested method, DNN-AM, outperforms the trained network in predicting ADHD in children, obtaining the lowest error rate, higher accuracy, and F1-Score. With both learning models, RF and DNN-AM the ADHD in children is diagnosed.