Exploring Heterogeneity in Neurodevelopmental Disorders Using EEG-based Normative Modeling
Poster No:
1646
Submission Type:
Abstract Submission
Authors:
Aida Ebadi1, Sahar Allouch1, Aya Kabbara1,2,3, Judie Tabbal4,1, Nadia Chabane5, Borja Rodríguez-Herreros5, Ahmad Mheich6, Mahmoud Hassan1
Institutions:
1MINDIG, Rennes, France, 2Lebanese Association for Scientific Research (LaSeR), Tripoli, Lebanon, 3Lebanese International University, Tripoli, Lebanon, 4Institut des Neurosciences Cliniques de Rennes (INCR), Rennes, France, 5CHUV Lausanne, Lausanne, Vaud, 6Lausanne University Hospital (CHUV), Lausanne, Vaud
First Author:
Co-Author(s):
Aya Kabbara
MINDIG|Lebanese Association for Scientific Research (LaSeR)|Lebanese International University
Rennes, France|Tripoli, Lebanon|Tripoli, Lebanon
MINDIG|Lebanese Association for Scientific Research (LaSeR)|Lebanese International University
Rennes, France|Tripoli, Lebanon|Tripoli, Lebanon
Judie Tabbal
Institut des Neurosciences Cliniques de Rennes (INCR)|MINDIG
Rennes, France|Rennes, France
Institut des Neurosciences Cliniques de Rennes (INCR)|MINDIG
Rennes, France|Rennes, France
Introduction:
Developing data-driven biomarkers in clinical neuroscience, particularly in psychiatry, faces major challenges due to the heterogeneous nature of mental disorders. Conventional case-control methods tend to overlook individual differences assuming homogeneity within groups and clear demarcations between groups. However, these assumptions do not accurately represent the complexity and the diversity within the population. To overcome this limitation, recent studies have embraced a more individualized, patient-centric approach using Normative Modeling (NM). It involves estimating reference trajectories for a brain phenotype and measuring individual deviations. NM's focus on individual rather than average patient profiles allows for the exploration of potential overlaps in brain activity between healthy individuals and patients, as well as among patients with different disorders. In this study, we explore the potential of using EEG-based normative models to investigate heterogeneity in Autism Spectrum Disorder.
Methods:
The study pipeline involves several key steps: EEG preprocessing, extraction of power spectral features, normative modeling, and subsequent analysis of the results.
We used resting state EEG from the Healthy Brain Network initiative led by the Child Mind Institute (N=3055, age: [5-22]). The dataset included a myriad of neurodevelopmental disorders with the majority being Attention deficit hyperactivity disorder (ADHD), Autism Spectrum Disorder (ASD), and Anxiety Disorder. EEG preprocessing and artifact removal were accomplished using a fully automated algorithm. The Power Spectrum Density (PSD) for each channel was computed using Welch's method. Generalized Additive Models for Location, Scale, and Shape (GAMLSS) models were fitted to the relative power of each channel and frequency band (delta, theta, alpha 1, alpha 2, beta, gamma).
Afterward, using an independent dataset (age: [5-10]) consisting of 13 subjects diagnosed with ASD and 31 Typically Developing (TD) subjects, we computed the deviation scores (Z-scores) from the reference trajectory for each individual (Figure 1). The Z-scores were further condensed for each group (TD and ASD) into percentages of positive/negative extremely deviant subjects (Z<-2, Z>2) per EEG channel (Figure 2).
We used resting state EEG from the Healthy Brain Network initiative led by the Child Mind Institute (N=3055, age: [5-22]). The dataset included a myriad of neurodevelopmental disorders with the majority being Attention deficit hyperactivity disorder (ADHD), Autism Spectrum Disorder (ASD), and Anxiety Disorder. EEG preprocessing and artifact removal were accomplished using a fully automated algorithm. The Power Spectrum Density (PSD) for each channel was computed using Welch's method. Generalized Additive Models for Location, Scale, and Shape (GAMLSS) models were fitted to the relative power of each channel and frequency band (delta, theta, alpha 1, alpha 2, beta, gamma).
Afterward, using an independent dataset (age: [5-10]) consisting of 13 subjects diagnosed with ASD and 31 Typically Developing (TD) subjects, we computed the deviation scores (Z-scores) from the reference trajectory for each individual (Figure 1). The Z-scores were further condensed for each group (TD and ASD) into percentages of positive/negative extremely deviant subjects (Z<-2, Z>2) per EEG channel (Figure 2).
Results:
Figure 2 shows higher percentages in TD as compared to ASD across the delta, alpha 2, beta, and gamma frequency bands. Notably, within the TD group, positive deviations in TD are consistently more prevalent across subjects in alpha 2, beta, and gamma bands, while negative deviations appear to exhibit higher consistency across subjects in the delta band. Furthermore, the percentage of extreme deviations in the theta band seems to be higher for the ASD compared to the TD group.
Conclusions:
The percentage of subjects with extreme regional deviations summarizes the consistency of individual regional deviations across TD and ASD cohorts. In the delta, alpha 2, beta, and gamma, no more than 15% of the individuals in the TD group deviate from the normative trajectory of the patients' group. In contrast, approximately 85% of the subjects closely align with the reference patient cohort. This highlights the nuanced difference between healthy controls and patients and provides evidence that case-control studies may be flawed when making significant distinctions between groups. Moreover, the low percentages of subjects exhibiting extreme regional deviations reflect the non-consistency of deviation topographical patterns across subjects, hence, corroborating the notion of heterogeneity within clinical and healthy cohorts.
To summarize, we developed normative curves for the relative power in EEG and demonstrated the possibility of quantifying the heterogeneity of neurodevelopmental disorders using NM.
To summarize, we developed normative curves for the relative power in EEG and demonstrated the possibility of quantifying the heterogeneity of neurodevelopmental disorders using NM.
Disorders of the Nervous System:
Neurodevelopmental/ Early Life (eg. ADHD, autism) 2
Modeling and Analysis Methods:
EEG/MEG Modeling and Analysis 1
Keywords:
Autism
Electroencephaolography (EEG)
Other - Normative Modeling, Heterogeneity
1|2Indicates the priority used for review
Provide references using author date format
Bethlehem, R. A. I., et al. 2022. "Brain charts for the human lifespan." Nature, 604(7906), 525–533.
Marquand, A. F., et al. 2019. "Conceptualizing mental disorders as deviations from normative functioning." Molecular Psychiatry, 24(10), 1415–1424.
Marquand, A. F., et al. 2016. "Understanding heterogeneity in clinical cohorts using normative models: beyond case-control studies." Biological Psychiatry, 80(7), 552–561.
Rutherford, S., et al. 2022. "Charting brain growth and aging at high spatial precision." eLife, 11, e72904.
Rutherford, S., et al. 2022. "The normative modeling framework for computational psychiatry." Nature Protocols, 17(7), 1711–1734.
Stasinopoulos, D. M., & Rigby, R. A. 2008. "Generalized additive models for location scale and shape (GAMLSS) in R." Journal of Statistical Software, 23, 1–46.
Verdi, S., et al. 2021. "Beyond the average patient: how neuroimaging models can address heterogeneity in dementia." Brain, 144(10), 2946–2953.