Predictors of real-time fMRI neurofeedback performance and improvement – A machine learning mega-analysis

Amelie Haugg,Kathrin Cohen Kadosh,Ruth A Lanius,Catharina Zich,Lydia Hellrung,Renate Schweizer,Frank Scharnowski,Sven Haller,Tibor Auer,Talma Hendler,Ronald Sladky,David Steyrl,Nikolaus Weiskopf,Marina Papoutsi,Dimitri Van De Ville,Simon H Kohl,Shuxia Yao,Tabea Kamp,Benjamin Becker,Jeff Macinnes,Bettina Sorger,Kirsten Emmert,Stavros Skouras,Kymberly Young,Fukuda Megumi,Theo Marins,Marcus Herdener,Amalia Mcdonald,Jerzy Bodurka,Kathryn C Dickerson,Fabian M Renz,Michael Marxen,Andrew A Nicholson,R Alison Adcock,Ralf Veit,Yury Koush,Manfred Hallschmid,Cameron Craddock,Matthias Kirschner,Gustavo Santo Pedro Pamplona ,Nan‐Kuei Chen ,Sook Lei Liew ,Cindy Sumaly Lor ,Jackob Nimrod Keynan ,Dong Youl Kim ,Jong Hwan Lee ,Maartje S Spetter ,Sebastian Jakob Götzendorfer

doi:10.1016/j.neuroimage.2021.118207

Abstract

Real-time fMRI neurofeedback is an increasingly popular neuroimaging technique that allows an individual to gain control over his/her own brain signals, which can lead to improvements in behavior in healthy participants as well as to improvements of clinical symptoms in patient populations. However, a considerably large ratio of participants undergoing neurofeedback training do not learn to control their own brain signals and, consequently, do not benefit from neurofeedback interventions, which limits clinical efficacy of neurofeedback interventions. As neurofeedback success varies between studies and participants, it is important to identify factors that might influence neurofeedback success. Here, for the first time, we employed a big data machine learning approach to investigate the influence of 20 different design-specific (e.g. activity vs. connectivity feedback), region of interest-specific (e.g. cortical vs. subcortical) and subject-specific factors (e.g. age) on neurofeedback performance and improvement in 608 participants from 28 independent experiments.With a classification accuracy of 60% (considerably different from chance level), we identified two factors that significantly influenced neurofeedback performance: Both the inclusion of a pre-training no-feedback run before neurofeedback training and neurofeedback training of patients as compared to healthy participants were associated with better neurofeedback performance. The positive effect of pre-training no-feedback runs on neurofeedback performance might be due to the familiarization of participants with the neurofeedback setup and the mental imagery task before neurofeedback training runs. Better performance of patients as compared to healthy participants might be driven by higher motivation of patients, higher ranges for the regulation of dysfunctional brain signals, or a more extensive piloting of clinical experimental paradigms. Due to the large heterogeneity of our dataset, these findings likely generalize across neurofeedback studies, thus providing guidance for designing more efficient neurofeedback studies specifically for improving clinical neurofeedback-based interventions. To facilitate the development of data-driven recommendations for specific design details and subpopulations the field would benefit from stronger engagement in open science research practices and data sharing.

Highlights

Real-time functional magnetic resonance neurofeedback is a non-invasive technique that enables healthy individuals and patients to voluntarily regulate neural signals
We observed that 69.41% of all participants were labelled as successful, meaning that for them, more than 50% of all neurofeedback training runs were successful
Our classification model achieved an accuracy of 60.3%, which was significantly better than chance level

Summary

Introduction

Real-time functional magnetic resonance (fMRI) neurofeedback is a non-invasive technique that enables healthy individuals and patients to voluntarily regulate neural signals. This method has gained growing popularity in the neuroimaging community and, to date, a wide range of real-time fMRI neurofeedback studies have collectively demonstrated the feasibility of volitional regulation through realtime fMRI neurofeedback (see Thibault et al (2018)) Many of these studies have shown behavioral changes in healthy individuals, as well as clinical improvements in patient populations after neurofeedback training. DeBettencourt et al, 2015; Pamplona et al, 2020), emotion regulation (Koush et al, 2015; Paret and Hendler, 2020; Zich et al, 2020), memory (e.g. Scharnowski et al, 2015; Sherwood et al, 2016; Zhang et al, 2013), motivation (e.g. Zhi et al, 2018), motor performance 2013), anxiety (Morgenroth et al, 2020), borderline personality disorder (Paret et al, 2016), depression (Linden et al, 2012; Mehler et al, 2018; Quevedo et al, 2020; Young et al, 2017, 2014), obsessive compulsive disorder (Buyukturkoglu et al, 2015), phobia (Zilverstand et al, 2015), post-traumatic stress disorder (Gerin et al, 2016; Nicholson et al, 2017), schizophrenia (Bauer et al, 2020), Tourette syndrome (Sukhodolsky et al, 2020), chronic pain (deCharms et al, 2005; Guan et al, 2014), Huntington’s disease (Papoutsi et al, 2018), obesity (Frank et al, 2012; Kohl et al, 2019), Parkinson’s disease (Buyukturkoglu et al, 2013; Subramanian et al, 2011), stroke rehabilitation (Mehler et al, 2020), tinnitus (Emmert et al, 2017; Haller et al, 2010), and visuo-spatial neglect (Fabien Robineau et al, 2019)

Methods

Results

Discussion

Conclusion