Distinct Methylphenidate-Evoked Response Measured Using Functional Near-Infrared Spectroscopy During Go/No-Go Task as a Supporting Differential Diagnostic Tool Between Attention-Deficit/Hyperactivity Disorder and Autism Spectrum Disorder Comorbid Children

Abstract

Attention deficit/hyperactivity disorder (ADHD) has been frequently reported as co-occurring with autism spectrum disorder (ASD). However, ASD-comorbid ADHD is difficult to diagnose since clinically significant symptoms are similar in both disorders. Therefore, we propose a classification method of differentially recognizing the ASD-comorbid condition in ADHD children. The classification method was investigated based on functional brain imaging measured by near-infrared spectroscopy (NIRS) during a go/no-go task. Optimization and cross-validation of the classification method was carried out in medicated-naïve and methylphenidate (MPH) administered ADHD and ASD-comorbid ADHD children (randomized, double-blind, placebo-controlled, and crossover design) to select robust parameters and cut-off thresholds. The parameters could be defined as either single or averaged multi-channel task-evoked activations under an administration condition (i.e., pre-medication, post-MPH, and post-placebo). The ADHD children were distinguished by significantly high MPH-evoked activation in the right hemisphere near the midline vertex. The ASD-comorbid ADHD children tended to have low activation responses in all regions. High specificity (86 ± 4.1%; mean ± SD), sensitivity (93 ± 7.3%), and accuracy (82 ± 1.6%) were obtained using the activation of oxygenated-hemoglobin concentration change in right middle frontal, angular, and precentral gyri under MPH medication. Therefore, the significantly differing MPH-evoked responses are potentially effective features and as supporting differential diagnostic tools.

Keywords: attention deficit/hyperactivity disorder; autism spectrum disorder; comorbid spectrum; differential diagnostic tool; inhibitory task-evoked activation; methylphenidate; near infra-red spectroscopy.

Figure 1

Spatial registration of fNIRS channels on bilateral hemispheric cortices.

Figure 2

Optimization operations (i.e., simple, OR, AND operations) for individual classification between ADHD and ASD-comorbid ADHD groups using either one- (i.e., simple operation); (A) or two-axis (i.e., OR and AND operations; (B,C) feature threshold. The estimation of ADHD classification is represented by patterned areas while ASD-comorbid ADHD subjects are categorized within plain areas.

Figure 3

Behavioral performances (accuracy – left y-axis and response time – right y-axis) of ADHD (gray-filled boxplots) and ASD-comorbid ADHD (i.e., void-filled boxplots) groups in the first measurement (A), post-MPH (B), and post-placebo (C) administrations. There was no significant difference between two groups in any performance data and administration conditions.

Figure 4

Statistical t-maps of ΔC_O2Hb and ΔC_HHb activation pattern in right hemisphere for both ADHD and ASD-comorbid ADHD children during GNG task before any administration (A), after medication (B), and placebo (C) administrations. The differences between ADHD and ASD-comorbid ADHD children were statistically examined (i.e., two-sample t-test). Two and single asterisks indicate channels with the significant activation by p < 0.01 and p < 0.05, respectively.

Figure 5

Activation coordinates for ADHD (black-dots) and ASD-comorbid ADHD (white-dots) groups using the optimum MPH-evoked response on ΔC_O2Hb activation in right MFG-ANG vs. right PrCG (A). Shaded regions are classification areas for the ASD-comorbid ADHD group using OR (gray-patch; cut-off thresholds at 0.04 mM⋅mm and 0.016 mM⋅mm for axis 1 and 2, respectively) and AND (magenta-patch; cut-off thresholds at 0.04 mM⋅mm and –0.15 mM⋅mm for axis 1 and 2, respectively) operations. Other classification operations are presented in red-, blue-, and black-plots for linear discriminant, quadratic discriminant, and SVM, respectively. ROC graphs of leave-one-out cross-validation results with the optimum MPH-evoked response on ΔC_O2Hb activation in right MFG-ANG vs. right PrCG (B) using OR (gray-plots) and AND (magenta-plots) operations. The bold lines indicate averages of cross-validation result. The shaded regions represent the range of validation performance (minimum–to-maximum specificity and sensitivity).

Figure 6

Subject-average for channel-wise ΔC_O2Hb (red- and magenta-plots) and ΔC_HHb (blue- and cyan-plots) waveforms for ADHD (N = 21; red- and blue-plots) and ASD-comorbid ADHD (N = 11; magenta- and cyan-plots) groups in right MFG (channels 32, 37, 41), right ANG (channel 44), and right PrCG (channel 42) regions. Patches around bold plots indicate standard error and gray-shaded interval is the stimulus interval of GNG task (24 s).

Figure 7

Performance differences between the two- (gray-filled boxplots) and three-axis (void-filled boxplots) feature thresholds using linear discriminant, quadratic discriminant, and SVM operations for specificity and sensitivity parameters. ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗∗p < 0.0001 for paired-sample t-test.