Normalization and deep learning based attention deficit hyperactivity disorder classification

Abstract

Attention Deficit Hyperactivity Disorder (ADHD) is one of the major mental-health disorders worldwide. ADHD is typically characterized by impaired executive function, impulsivity, hyperactivity and with respect to these behavioral symptoms, diagnosis of ADHD is performed. These symptoms are obviously seen at in early stage. Serious impairments and substantial burdens are induced for society as well as to families. However, for ADHD, there is no diagnostic laboratory in current scenario. Psychological tests like Brown Attention Deficit Disorder Scale (BADDS), Conners Parent Rating Scale and ADHD Rating Scale (ADHD-RS) are carried out for ADHD diagnosis. Tedious and complex clinical analysis are needed in this testing and this makes low efficiency of the diagnostic process. A traditional diagnosis technique of ADHD produces degraded results. So, enhanced extreme learning machine is incorporated with existing techniques for avoiding the issues of performance degradation. There is a need to enhance the classifier performance further and there is a chance for unwanted noise in input samples, which may degrade the performance of classifier. For avoiding these issues, an enhanced and automated ADHS diagnosis technique is proposed. First stage is pre-processing, and it is carried out based on min max normalization and feature extraction is a next stage, which is carried out through Fast Independent Component Analysis and third stage is a Deep Extreme Learning Machine (DELM) based ADHD identification and classification. Extreme Learning Machine with Kernel (KELM) and Multilayer Extreme Learning Machine (MLELM) algorithm are combined in this method and it is termed as deep extreme learning machine (DELM). Collection of neuro images are used for quantitative and qualitative analysis and with respect to f-measure, recall, precision and accuracy, robustness of proposed technique is demonstrated.