Examining cognition and brain networks using magnetoencephalography in paediatric autoimmune encephalitis and acute disseminated encephalomyelitis: a preliminary study

Abstract

Paediatric autoimmune encephalitis, including acute disseminated encephalomyelitis, are inflammatory brain diseases presenting with cognitive deficits, psychiatric symptoms, seizures, MRI and EEG abnormalities. Despite improvements in disease recognition and early immunotherapy, long-term outcomes in paediatric autoimmune encephalitis remain poor. Our aim was to understand functional connectivity changes that could be associated with negative developmental outcomes across different types of paediatric autoimmune encephalitis using magnetoencephalography. Participants were children diagnosed with paediatric autoimmune encephalitis at least 18 months before testing and typically developing children. All completed magnetoencephalography recording at rest, T₁ MRI scans and neuropsychology testing. Brain connectivity (specifically in delta and theta) was estimated with amplitude envelope correlation, and network efficiency was measured using graph measures (global efficiency, local efficiency and modularity). Twelve children with paediatric autoimmune encephalitis (11.2 ± 3.5 years, interquartile range 9 years; 5M:7F) and 12 typically developing controls (10.6 ± 3.2 years, interquartile range 7 years; 8M:4F) participated. Children with paediatric autoimmune encephalitis did not differ from controls in working memory (t(21) = 1.449; P = 0.162; d = 0.605) but had significantly lower processing speed (t(21) = 2.463; P = 0.023; Cohen's d = 1.028). Groups did not differ in theta network topology measures. The paediatric autoimmune encephalitis group had a significantly lower delta local efficiency across all thresholds tested (d = -1.60 at network threshold 14%). Theta modularity was associated with lower working memory (β = -0.781; t(8) = -2.588, P = 0.032); this effect did not survive correction for multiple comparisons (P(corr) = 0.224). Magnetoencephalography was able to capture specific network alterations in paediatric autoimmune encephalitis patients. This preliminary study demonstrates that magnetoencephalography is an appropriate tool for assessing children with paediatric autoimmune encephalitis and could be associated with cognitive outcomes.

Keywords: NMDAR-Ab encephalitis; acute disseminated encephalomyelitis (ADEM); autoimmune encephalitis; magnetoencephalography (MEG); neurodevelopment.

Graphical Abstract

Figure 1

Cognitive outcome from the neuropsychological assessments across groups. Group comparisons were tested using independent t-tests. The AE group had a significantly lower average score in processing speed [t(21) = 2.463; *P = 0.023; Cohen’s d = 1.028]. Groups did not significantly differ in working memory [t(21) = 1.449; P = 0.162; Cohen’s d = 0.605]. AE, autoimmune encephalitis; PSI, processing speed index; WMI, working memory index; measures were obtained with the WISC-V.

Figure 2

Cognitive scores categorized according to the normative WISC-V classification. AE, autoimmune encephalitis. Classification categories and measures were obtained with the WISC-V.

Figure 3

Comparison of delta network measures (1–4 Hz) between AE and control groups, showing a significantly lower local efficiency in AE. S.mtpc, maximum observed statistic; S.crit, critical value of the null maximum statistic; A.crit, mean of the supracritical null AUCs; A.mtpc, AUC value of supracritical cluster. Each graph depicts the network contrast of the AE group in reference to the control group. The dots are the maximum null t statistics of the 5000 permutations. Dots where the thick line is crossing are the observed t statistics. The dashed line is the critical null statistic (S.crit, top 95th percentile of the null distribution of maximum t statistics). The shaded areas represent clusters of observed statistics above the critical value, whose area-under-the-curve (A.mtpc) is also greater than the mean areas-under-the-curve of the supracritical permuted statistics (A.crit). This means that a lower shaded bar is a significantly lower network measure compared with the control group (at P < 0.05). Non-shaded bars are non-significant. The reported effect size d is only that of the threshold with the largest difference.

Figure 4

Comparison of theta network measures (5–8 Hz) between AE and control groups, showing no significant differences S.mtpc, maximum observed statistic; S.crit, critical value of the null max. statistic; A.crit, mean of the supracritical null AUCs; A.mtpc, AUC value of supracritical cluster. Each graph depicts the network contrast of the AE group in reference to the control group. The dots are the maximum null t statistics of the 5000 permutations. Dots where the thick line is crossing are the observed t statistics. The dashed line is the critical null statistic (S.crit, top 95th percentile of the null distribution of maximum t statistics). The shaded areas represent clusters of observed statistics above the critical value, whose area-under-the-curve (A.mtpc) is also greater than the mean areas-under-the-curve of the supracritical permuted statistics (A.crit). This means that a lower shaded bar is a significantly lower network measure compared with the control group (at P < 0.05). Non-shaded bars are non-significant. The reported effect size d is only that of the threshold with the largest difference.