MENU

Concept2Brain: AI-driven EEG simulation from images and text

PsyPost USA
Overview
Researchers have developed Concept2Brain, an open-access AI tool that generates synthetic electroencephalography (EEG) recordings based on photographs and written text. By simulating human brain activity during visual and emotional perception, the system allows researchers to conduct preliminary experiments on computers before recruiting human volunteers.
In Depth

Key Findings

Concept2Brain is a generative AI tool designed to predict how human brain waves respond to specific visual or textual stimuli. Published in Nature Communications, the research demonstrates that the platform can produce synthetic EEG recordings that resemble real human brain activity. The system effectively bridges the gap between conceptual meaning and scalp electrical activity, allowing for the simulation of brain-wave experiments without the immediate need for costly and time-consuming human data collection.

Technical Details

The architecture consists of two linked deep neural networks:

  • Conditional Variational Autoencoder: This network compresses 129-channel EEG recordings into a compact summary of 256 numbers. It includes a participant identifier to preserve individual differences in brain-wave shapes.
  • Translation Network: Built around OpenAI’s CLIP system, this network maps images and text into a shared 512-number conceptual space. A three-layer neural network then translates these conceptual numbers into the 256-number brain-wave summary.

To train the model, researchers used EEG data from 88 university volunteers across two experiments. Participants viewed 360 unique photographs—divided equally among pleasant, neutral, and unpleasant scenes—recorded via 128 scalp sensors and one reference sensor at 500 samples per second. After filtering noise, the team utilized 13,640 clean 1.5-second trials.

Background & Context

While AI has previously been used for “decoding” (reconstructing meaning from brain scans), Concept2Brain focuses on “encoding”—predicting electrical waves from a starting image or phrase. This builds on previous research, including a 2023 multimodal project cataloging 1,854 object categories and a 2008 study linking emotional arousal to the magnitude of electrical brain responses.

Significance & Outlook

Lead author Alejandro Santos-Mayo states that the tool provides a non-linear, multivariate approach to predicting brain data. By offering the model’s code, trained weights, and a public website, the researchers aim to provide a free, open-access resource for simulating EEG activity and deepening the understanding of neurophysiological components.

Source: https://www.psypost.org/scientists-built-an-ai-called-concept2brain-that-simulates-human-brain-waves-from-pictures-and-text/

Get our weekly technology intelligence — free

Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.

Subscribe Free — Weekly Tech Intelligence

By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.

  • Your email and selected fields are used only to deliver the newsletter.
  • We never share your information with third parties.
  • You can unsubscribe anytime via the link in each email.

See our Privacy Policy for details.

Takes about a minute · Unsubscribe anytime

Published by Troy-Technical, an independent site run by one engineer with a career in materials development.
About the author / Contact info@troy-technical.jp
Let's share this post !

Author of this article

TOC