Diurnal variation of motor activity in adult ADHD patients analyzed with methods from graph theory

Abstract

Attention-deficit /hyperactivity disorder (ADHD) is a common neurodevelopmental syndrome characterized by age-inappropriate levels of motor activity, impulsivity and attention. The aim of the present study was to study diurnal variation of motor activity in adult ADHD patients, compared to healthy controls and clinical controls with mood and anxiety disorders. Wrist-worn actigraphs were used to record motor activity in a sample of 81 patients and 30 healthy controls. Time series from registrations in the morning and evening were analyzed using measures of variability, complexity and a newly developed method, the similarity algorithm, based on transforming time series into graphs. In healthy controls the evening registrations showed higher variability and lower complexity compared to morning registrations, however this was evident only in the female controls. In the two patient groups the same measures were not significantly different, with one exception, the graph measure bridges. This was the measure that most clearly separated morning and evening registrations and was significantly different both in healthy controls and in patients with a diagnosis of ADHD. These findings suggest that actigraph registrations, combined with mathematical methods based on graph theory, may be used to elucidate the mechanisms responsible for the diurnal regulation of motor activity.

Complexity and Variability Analyses of Motor Activity Distinguish Mood States in Bipolar Disorder

“…When changing the values of k , different similarity graphs are obtained, and different results are to be expected. 23 However, the results of the k = 5 neighborhood analyses were quite similar to the k = 2 results for both the 20-hour and the 120-minute evening sequences. The morning sequences only resulted in differences in edges and missing edges.…”

“…In this paper, we apply the nonlinear similarity graph algorithm which is based on work done by Lacasa et al, 34 and has been comprehensively described. 22,23 This algorithm transforms a time series S=(x 1 , x 2 , … x n ) into an undirected similarity graph G. Each element of time series S corresponds to a node u in V= {1, 2, … n} and each node u is assigned a weight equal to the value of x u. The distance between two nodes u and v is |u-v| and when the distance is 1, the two nodes u and v are defined as direct neighbors.…”

“…Defining 20 % as the threshold for similarity is founded in previous studies of motor activity with the sample entropy method, 9,10 as well as applied in our groups’ previous studies of motor activity and the similarity graph algorithm. 22,23 We used a selection of similarity graph parameters to analyze our data. In summary, any node of the graph with few or no neighbors indicates an alteration in activity.…”

“…Bridges expose more subtle activity fluctuations. 23 The number of cliques represents the smoothness of activity fluctuations in a time series. 35 We report on 3-cliques.…”

“…The method has successfully differentiated between depression, schizophrenia and healthy controls, 22 as well as patients with ADHD from clinical controls (depression/anxiety) and healthy controls. 23…”