Preparatory activity of anterior insula predicts conflict errors: integrating convolutional neural networks and neural mass models

Kaboodvand, Neda; Karimi, Hanie; Iravani, Behzad · 2024 · DOAJ

DOI: 10.1038/s41598-024-67034-5

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates the neural mechanisms of proactive cognitive control, specifically examining how pre-stimulus preparatory electrophysiological activity (PEA) predicts performance in conflict-inducing tasks. While functional neuroimaging has identified frontal structures involved in cognitive control, the low temporal resolution of fMRI and the limitations of traditional univariate EEG analyses have hindered a precise understanding of the millisecond-scale dynamics underlying proactive control. The authors aimed to bridge this gap by integrating deep learning with biologically inspired neural modeling to identify predictive biomarkers and test causal interventions. The researchers utilized two open-access datasets: concurrent EEG-fMRI data from participants performing Stroop and Simon tasks, and psychophysiological-fMRI data from a neurofeedback study involving volitional emotional control. They developed a neuroimaging-informed convolutional neural network (CNN) to predict conflict errors (Hits vs. Errors) based on PEA from 19 regions of interest identified via fMRI. To identify the most critical regions, they employed perturbation-based occlusion sensitivity analysis. Furthermore, they integrated the CNN with a Jansen-Rit neural mass model to simulate the effects of high-frequency neurostimulation on cognitive performance, validating these theoretical findings against experimental data on intrinsic neuromodulation. The CNN successfully predicted conflict errors with 64% accuracy, significantly above chance. Analysis revealed that predictive PEA was characterized by arrhythmic broadband neural dynamics rather than specific frequency bands. Occlusion experiments identified the right anterior insula, right precentral gyrus, and right pars opercularis as the most predictive regions for prospective cognitive control. However, simulations using the neural mass model demonstrated that high-frequency stimulation (130 Hz) of the *left* anterior insula significantly enhanced cognitive performance and reduced conflict errors, despite the right anterior insula having higher predictive value. Experimental validation using neurofeedback data confirmed that the left anterior insula is intrinsically modulated during volitional control of emotional valence, but not arousal. These findings highlight the central role of the anterior insula in orchestrating proactive cognitive control. Crucially, the study demonstrates a dissociation between regions serving as biomarkers for predictive activity and those serving as effective targets for neurostimulation. This distinction suggests that while the right anterior insula reflects readiness for conflict detection, stimulating the left anterior insula may be more effective for enhancing cognitive control. The integration of CNNs with neural mass models provides a robust framework for identifying causal mechanisms and optimizing closed-loop neuromodulation interventions.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• executive function