Leveraging Machine Learning Approaches for Predicting Antidepressant Treatment Response Using Electroencephalography (EEG) and Clinical Data

Abstract

Background: Individuals with major depressive disorder (MDD) vary in their response to antidepressants. However, identifying objective biomarkers, prior to or early in the course of treatment that can predict antidepressant efficacy, remains a challenge.Methods: Individuals with MDD participated in a 12-week antidepressant pharmacotherapy trial. Electroencephalographic (EEG) data was collected before and 1 week post-treatment initiation in 51 patients. Response status at week 12 was established with the Montgomery-Asberg Depression Scale (MADRS), with a ≥50% decrease characterizing responders (N = 27/24 responders/non-responders). We used a machine learning (ML)-approach for predicting response status. We focused on Random Forests, though other ML methods were compared. First, we used a tree-based estimator to select a relatively small number of significant features from: (a) demographic/clinical data (age, sex, individual item/total MADRS scores at baseline, week 1, change scores); (b) scalp-level EEG power; (c) source-localized current density (via exact low-resolution electromagnetic tomography [eLORETA] software). Second, we applied kernel principal component analysis to reduce and map important features. Third, a set of ML models were constructed to classify response outcome based on mapped features. For each dataset, predictive features were extracted, followed by a model of all predictive features, and finally by a model of the most predictive features.Results: Fifty eLORETA features were predictive of response (across bands, both time-points); alpha1/theta eLORETA features showed the highest predictive value. Eighty-eight scalp EEG features were predictive of response (across bands, both time-points), with theta/alpha2 being most predictive. Clinical/demographic data consisted of 31 features, with the most important being week 1 “concentration difficulty” scores. When all features were included into one model, its predictive utility was high (88% accuracy). When the most important features were extracted in the final model, 12 predictive features emerged (78% accuracy), including baseline scalp-EEG frontopolar theta, parietal alpha2 and frontopolar alpha1.Conclusions: These findings suggest that ML models of pre- and early treatment-emergent EEG profiles and clinical features can serve as tools for predicting antidepressant response. While this must be replicated using large independent samples, it lays the groundwork for research on personalized, “biomarker”-based treatment approaches.

Frontal theta and posterior alpha in resting EEG: A critical examination of convergent and discriminant validity

“…Moreover, the alpha rhythm is prominent (e.g., visible in raw EEG traces) and reliably quantified by different research groups using different methodologies (e.g., Barry & De Blasio, 2018;Labounek et al, 2018;Schmidt et al, 2017;Shackman et al, 2010;Sockeel, Schwartz, Pélégrini-issac, & Benali, 2016;Tenke et al, 2017). Importantly, greater posterior alpha oscillations at rest predicted a favorable clinical outcome for individuals diagnosed with MDD (Baskaran et al, 2017;Bruder et al, 2008;Jaworska, de la Salle, Ibrahim, Blier, & Knott, 2019;Kandilarova et al, 2017;Knott et al, 1996;Tenke et al, 2011;Ulrich, Renfordt, Zeller, & Frick, 1984;Ulrich, Renfordt, & Frick, 1986; although see Arns et al, 2016, andKnott, Mahoney, Kennedy, &Evans, 2000, for unsuccessful attempts to replicate these findings). To some degree, these reports differed in methodology, including EEG montage (density, locations) and reference, preprocessing steps, and a priori selection of frequency bins.…”

“…This is important for clinical utility, as specific brain regions likely have differential relatedness to psychiatric outcomes, and EEG reference choices will vary from clinic to clinic. There are myriad techniques and tools available to accomplish these aims, yet it remains to be seen which specific techniques are best for clinical prediction (Bridwell et al, 2018;Delorme et al, 2012;Jaworska et al, 2019). Nonetheless, pivoting toward analyses focused on latent variables should substantially advance biomarker development and utility.…”

An electroencephalographic signature predicts antidepressant response in major depression

“…This heterogeneity may account for the modest superiority of antidepressant medication over placebo (Cohen’s d ~ 0.3) 2 - 6 . Work over the past two decades has suggested that resting-state electroencephalography (rsEEG) may be able to identify treatment-predictive heterogeneity in depression 7 - 10 . Specific attention has been paid to prefrontal and parietal signals carried by the theta (4-7Hz) and alpha (8-12Hz) frequency bands 7 - 13 .…”

Neuroimaging Biomarkers for Predicting Treatment Response and Recurrence of Major Depressive Disorder

“…In an additional recent study, machine learning models of EEG profiles and clinical features were predictive of the antidepressant response [72]; however, since the number of participants in some previous EEG studies is relatively small [68,69] and the results are inconsistent, EEG studies must be replicated using large independent samples in the future.…”