Inference of ADHD using linear discriminant feature analysis of brain caudate nucleus

Abstract

A methodology to inferAttention Deficit/Hyperactivity Disorder (ADHD) from the features of Caudate Nucleus using Linear Discriminant Analysis is proposed in this research work. A Brain Image Data Set containing 40, T2 axial Human ADHD Brain MR Images is created initially. The caudate Nucleus of each image is extracted with Fuzzy C-Means algorithm and stored in the Caudate Nucleus Data Set(CNDS). The texture, intensity and shape based features of CNDS is extracted, analysed and the features which are closely related to the inference of ADHD are identified. With the help of statistical measures and LDA classifier the input images are classified into ADHD and NOADHD classes and further the level of ADHD is identified as Mild, Moderate and severe ADHD images. The LDA classifier results show that out of 40CN brain tissues, 36ADHD images are found with reduced volume, contrast and intensity values compared to control group and 4 images are found with NoADHD. The performanceis evaluated by means of various classification metrics like Confusion matrix, Accuracy measures and Parallel Coordinates Plot. The LDA classifier yields90% of classification accuracy on CN dataset and reveals that the children with ADHD are found with low volume, contrast and Intensity values compared to normal group which suggests the neurologists to diagnose ADHD using Brain scan method instead of Medication or counseling.