| Attention deficit/hyperactivity disorder (ADHD) is a neurodevelopmental disorder that is typically diagnosed in young children. ADHD is heterogeneous by nature with subjects exhibiting various combinations of inattention, impulsiveness, and hyperactivity. ADHD typically persists into adulthood and increases the likelihood of diverse mental health issues and comorbid disorders such as depression, anxiety and learning disabilities. The ramifications of ADHD may worsen with age.In this paper, we propose a novel approach for the diagnosis of ADHD from resting-state fMRI (rs-fMRI) images using Capsule Network paired with LSTM Network. Combining the predictions of the Capsule Network along with the LSTM Network with the help of a voting classifier ensures that both aspects of the data - the sequential features for every subject’s scan from the LSTM Network along with the attributes of the entire scan itself from the Capsule Network can be combined to fill in some gaps between each of their predictions and give a better prediction as a whole. Our proposed model achieves an accuracy of 80% on the KKI dataset and 73.33% on Peking-I dataset which is an improvement over the existing approaches. |
*** Title, author list and abstract as seen in the Camera-Ready version of the paper that was provided to Conference Committee. Small changes that may have occurred during processing by Springer may not appear in this window.