MindSemantix: Deciphering Brain Visual Experiences with a Brain-Language Model

Deciphering the human visual experience through brain activities captured by fMRI represents a compelling and cutting-edge challenge in the field of neuroscience research. Compared to merely predicting the viewed image itself, decoding brain activity into meaningful captions provides a higher-level interpretation and summarization of visual information, which naturally enhances the application flexibility in real-world situations. In this work, we introduce MindSemantix, a novel multi-modal framework that enables LLMs to comprehend visually-evoked semantic content in brain activity. Our MindSemantix explores a more ideal brain captioning paradigm by weaving LLMs into brain activity analysis, crafting a seamless, end-to-end Brain-Language Model. To effectively capture semantic information from brain responses, we propose Brain-Text Transformer, utilizing a Brain Q-Former as its core architecture. It integrates a pre-trained brain encoder with a frozen LLM to achieve multi-modal alignment of brain-vision-language and establish a robust brain-language correspondence. To enhance the generalizability of neural representations, we pre-train our brain encoder on a large-scale, cross-subject fMRI dataset using self-supervised learning techniques. MindSemantix provides more feasibility to downstream brain decoding tasks such as stimulus reconstruction. Conditioned by MindSemantix captioning, our framework facilitates this process by integrating with advanced generative models like Stable Diffusion and excels in understanding brain visual perception. MindSemantix generates high-quality captions that are deeply rooted in the visual and semantic information derived from brain activity. This approach has demonstrated substantial quantitative improvements over prior art. Our code will be released.

Reconstructing Perceived Images from Brain Activity by Visually-guided Cognitive Representation and Adversarial Learning

Reconstructing perceived images based on brain signals measured with functional magnetic resonance imaging (fMRI) is a significant and meaningful task in brain-driven computer vision. However, the inconsistent distribution and representation between fMRI signals and visual images cause the heterogeneity gap, which makes it challenging to learn a reliable mapping between them. Moreover, considering that fMRI signals are extremely high-dimensional and contain a lot of visually-irrelevant information, effectively reducing the noise and encoding powerful visual representations for image reconstruction is also an open problem. We show that it is possible to overcome these challenges by learning a visually-relevant latent representation from fMRI signals guided by the corresponding visual features, and recovering the perceived images via adversarial learning. The resulting framework is called Dual-Variational Autoencoder/ Generative Adversarial Network (D-VAE/GAN). By using a novel 3-stage training strategy, it encodes both cognitive and visual features via a dual structure variational autoencoder (D-VAE) to adapt cognitive features to visual feature space, and then learns to reconstruct perceived images with generative adversarial network (GAN). Extensive experiments on three fMRI recording datasets show that D-VAE/GAN achieves more accurate visual reconstruction compared with the state-of-the-art methods.