Deep image reconstruction from human brain activity

Postmortem communication

The phenomenon of near death and dying experiences has been both of popular interest and of scientific speculation. However, the reality of mental perception at the point of death is currently a subjective experience and has not been formally evaluated. While postmortem gene expression, even in humans, has been evaluated, restoration of postmortem brain activity has heretofore only been attempted in animal models, at the molecular and cellular levels. Meanwhile, progress has been made to translate brain activity of living humans into speech and images. This paper proposes two inter-related thought experiments. First, assuming progress and refinement of the technology of translating human brain activity into interpretable speech and images, can an objective analysis of death experiences be obtained by utilizing these technologies on dying humans? Second, can human brain function be revived postmortem and, if so, can the relevant technologies be utilized for communication with (recently) deceased individuals? In this paper, these questions are considered and possible implications explored.

Factorized visual representations in the primate visual system and deep neural networks

Object classification has been proposed as a principal objective of the primate ventral visual stream and has been used as an optimization target for deep neural network models (DNNs) of the visual system. However, visual brain areas represent many different types of information, and optimizing for classification of object identity alone does not constrain how other information may be encoded in visual representations. Information about different scene parameters may be discarded altogether ('invariance'), represented in non-interfering subspaces of population activity ('factorization') or encoded in an entangled fashion. In this work, we provide evidence that factorization is a normative principle of biological visual representations. In the monkey ventral visual hierarchy, we found that factorization of object pose and background information from object identity increased in higher-level regions and strongly contributed to improving object identity decoding performance. We then conducted a large-scale analysis of factorization of individual scene parameters - lighting, background, camera viewpoint, and object pose - in a diverse library of DNN models of the visual system. Models which best matched neural, fMRI, and behavioral data from both monkeys and humans across 12 datasets tended to be those which factorized scene parameters most strongly. Notably, invariance to these parameters was not as consistently associated with matches to neural and behavioral data, suggesting that maintaining non-class information in factorized activity subspaces is often preferred to dropping it altogether. Thus, we propose that factorization of visual scene information is a widely used strategy in brains and DNN models thereof.

Large-scale foundation models and generative AI for BigData neuroscience

Recent advances in machine learning have led to revolutionary breakthroughs in computer games, image and natural language understanding, and scientific discovery. Foundation models and large-scale language models (LLMs) have recently achieved human-like intelligence thanks to BigData. With the help of self-supervised learning (SSL) and transfer learning, these models may potentially reshape the landscapes of neuroscience research and make a significant impact on the future. Here we present a mini-review on recent advances in foundation models and generative AI models as well as their applications in neuroscience, including natural language and speech, semantic memory, brain-machine interfaces (BMIs), and data augmentation. We argue that this paradigm-shift framework will open new avenues for many neuroscience research directions and discuss the accompanying challenges and opportunities.

Computational reconstruction of mental representations using human behavior

Revealing how the mind represents information is a longstanding goal of cognitive science. However, there is currently no framework for reconstructing the broad range of mental representations that humans possess. Here, we ask participants to indicate what they perceive in images made of random visual features in a deep neural network. We then infer associations between the semantic features of their responses and the visual features of the images. This allows us to reconstruct the mental representations of multiple visual concepts, both those supplied by participants and other concepts extrapolated from the same semantic space. We validate these reconstructions in separate participants and further generalize our approach to predict behavior for new stimuli and in a new task. Finally, we reconstruct the mental representations of individual observers and of a neural network. This framework enables a large-scale investigation of conceptual representations.

A New Framework Combining Diffusion Models and the Convolution Classifier for Generating Images from EEG Signals

The generation of images from electroencephalography (EEG) signals has become a popular research topic in recent research because it can bridge the gap between brain signals and visual stimuli and has wide application prospects in neuroscience and computer vision. However, due to the high complexity of EEG signals, the reconstruction of visual stimuli through EEG signals continues to pose a challenge. In this work, we propose an EEG-ConDiffusion framework that involves three stages: feature extraction, fine-tuning of the pretrained model, and image generation. In the EEG-ConDiffusion framework, classification features of EEG signals are first obtained through the feature extraction block. Then, the classification features are taken as conditions to fine-tune the stable diffusion model in the image generation block to generate images with corresponding semantics. This framework combines EEG classification and image generation means to enhance the quality of generated images. Our proposed framework was tested on an EEG-based visual classification dataset. The performance of our framework is measured by classification accuracy, 50-way top-k accuracy, and inception score. The results indicate that the proposed EEG-Condiffusion framework can extract effective classification features and generate high-quality images from EEG signals to realize EEG-to-image conversion.

Uncovering the Role of the Early Visual Cortex in Visual Mental Imagery

The question of whether the early visual cortex (EVC) is involved in visual mental imagery remains a topic of debate. In this paper, I propose that the inconsistency in findings can be explained by the unique challenges associated with investigating EVC activity during imagery. During perception, the EVC processes low-level features, which means that activity is highly sensitive to variation in visual details. If the EVC has the same role during visual mental imagery, any change in the visual details of the mental image would lead to corresponding changes in EVC activity. Within this context, the question should not be whether the EVC is 'active' during imagery but how its activity relates to specific imagery properties. Studies using methods that are sensitive to variation in low-level features reveal that imagery can recruit the EVC in similar ways as perception. However, not all mental images contain a high level of visual details. Therefore, I end by considering a more nuanced view, which states that imagery can recruit the EVC, but that does not mean that it always does so.

Brain2GAN: Feature-disentangled neural encoding and decoding of visual perception in the primate brain

A challenging goal of neural coding is to characterize the neural representations underlying visual perception. To this end, multi-unit activity (MUA) of macaque visual cortex was recorded in a passive fixation task upon presentation of faces and natural images. We analyzed the relationship between MUA and latent representations of state-of-the-art deep generative models, including the conventional and feature-disentangled representations of generative adversarial networks (GANs) (i.e., z- and w-latents of StyleGAN, respectively) and language-contrastive representations of latent diffusion networks (i.e., CLIP-latents of Stable Diffusion). A mass univariate neural encoding analysis of the latent representations showed that feature-disentangled w representations outperform both z and CLIP representations in explaining neural responses. Further, w-latent features were found to be positioned at the higher end of the complexity gradient which indicates that they capture visual information relevant to high-level neural activity. Subsequently, a multivariate neural decoding analysis of the feature-disentangled representations resulted in state-of-the-art spatiotemporal reconstructions of visual perception. Taken together, our results not only highlight the important role of feature-disentanglement in shaping high-level neural representations underlying visual perception but also serve as an important benchmark for the future of neural coding.

Achieving more human brain-like vision via human EEG representational alignment

Despite advancements in artificial intelligence, object recognition models still lag behind in emulating visual information processing in human brains. Recent studies have highlighted the potential of using neural data to mimic brain processing; however, these often rely on invasive neural recordings from non-human subjects, leaving a critical gap in understanding human visual perception. Addressing this gap, we present, for the first time, 'Re(presentational)Al(ignment)net', a vision model aligned with human brain activity based on non-invasive EEG, demonstrating a significantly higher similarity to human brain representations. Our innovative image-to-brain multi-layer encoding framework advances human neural alignment by optimizing multiple model layers and enabling the model to efficiently learn and mimic human brain's visual representational patterns across object categories and different modalities. Our findings suggest that ReAlnet represents a breakthrough in bridging the gap between artificial and human vision, and paving the way for more brain-like artificial intelligence systems.

May I see what you see? Predicting visual features from neuronal activity

Understanding brain response to audiovisual stimuli is a key challenge in understanding neuronal processes. In this paper, we describe our effort aimed at reconstructing video frames from observed functional MRI images. We also demonstrate that our model can predict visual objects. Our method constructs an autoencoder model for a set of training video segments to code video streams into their corresponding latent representations. Next, we learn a mapping from the observed fMRI response to the corresponding latent video frame representation. Finally, we pass the latent vectors computed using the fMRI response through the decoder to reconstruct the predicted image. We show that the representations of video frames and those constructed from corresponding fMRI images are highly clustered, the latent representations can be used to predict objects in video frames using just the fMRI frames, and fMRI responses can be used to reconstruct the inputs to predict the presence of faces.

Mental image reconstruction from human brain activity: Neural decoding of mental imagery via deep neural network-based Bayesian estimation

Visual images observed by humans can be reconstructed from their brain activity. However, the visualization (externalization) of mental imagery is challenging. Only a few studies have reported successful visualization of mental imagery, and their visualizable images have been limited to specific domains such as human faces or alphabetical letters. Therefore, visualizing mental imagery for arbitrary natural images stands as a significant milestone. In this study, we achieved this by enhancing a previous method. Specifically, we demonstrated that the visual image reconstruction method proposed in the seminal study by Shen et al. (2019) heavily relied on low-level visual information decoded from the brain and could not efficiently utilize the semantic information that would be recruited during mental imagery. To address this limitation, we extended the previous method to a Bayesian estimation framework and introduced the assistance of semantic information into it. Our proposed framework successfully reconstructed both seen images (i.e., those observed by the human eye) and imagined images from brain activity. Quantitative evaluation showed that our framework could identify seen and imagined images highly accurately compared to the chance accuracy (seen: 90.7%, imagery: 75.6%, chance accuracy: 50.0%). In contrast, the previous method could only identify seen images (seen: 64.3%, imagery: 50.4%). These results suggest that our framework would provide a unique tool for directly investigating the subjective contents of the brain such as illusions, hallucinations, and dreams.

Brain encoding models based on multimodal transformers can transfer across language and vision

Encoding models have been used to assess how the human brain represents concepts in language and vision. While language and vision rely on similar concept representations, current encoding models are typically trained and tested on brain responses to each modality in isolation. Recent advances in multimodal pretraining have produced transformers that can extract aligned representations of concepts in language and vision. In this work, we used representations from multimodal transformers to train encoding models that can transfer across fMRI responses to stories and movies. We found that encoding models trained on brain responses to one modality can successfully predict brain responses to the other modality, particularly in cortical regions that represent conceptual meaning. Further analysis of these encoding models revealed shared semantic dimensions that underlie concept representations in language and vision. Comparing encoding models trained using representations from multimodal and unimodal transformers, we found that multimodal transformers learn more aligned representations of concepts in language and vision. Our results demonstrate how multimodal transformers can provide insights into the brain's capacity for multimodal processing.

Reconstructing visual illusory experiences from human brain activity

Visual illusions provide valuable insights into the brain's interpretation of the world given sensory inputs. However, the precise manner in which brain activity translates into illusory experiences remains largely unknown. Here, we leverage a brain decoding technique combined with deep neural network (DNN) representations to reconstruct illusory percepts as images from brain activity. The reconstruction model was trained on natural images to establish a link between brain activity and perceptual features and then tested on two types of illusions: illusory lines and neon color spreading. Reconstructions revealed lines and colors consistent with illusory experiences, which varied across the source visual cortical areas. This framework offers a way to materialize subjective experiences, shedding light on the brain's internal representations of the world.

Adversarial attacks and defenses using feature-space stochasticity

Recent studies in deep neural networks have shown that injecting random noise in the input layer of the networks contributes towards ℓp-norm-bounded adversarial perturbations. However, to defend against unrestricted adversarial examples, most of which are not ℓp-norm-bounded in the input layer, such input-layer random noise may not be sufficient. In the first part of this study, we generated a novel class of unrestricted adversarial examples termed feature-space adversarial examples. These examples are far from the original data in the input space but adjacent to the original data in a hidden-layer feature space and far again in the output layer. In the second part of this study, we empirically showed that while injecting random noise in the input layer was unable to defend these feature-space adversarial examples, they were defended by injecting random noise in the hidden layer. These results highlight the novel benefit of stochasticity in higher layers, in that it is useful for defending against these feature-space adversarial examples, a class of unrestricted adversarial examples.

Dual-Guided Brain Diffusion Model: Natural Image Reconstruction from Human Visual Stimulus fMRI

The reconstruction of visual stimuli from fMRI signals, which record brain activity, is a challenging task with crucial research value in the fields of neuroscience and machine learning. Previous studies tend to emphasize reconstructing pixel-level features (contours, colors, etc.) or semantic features (object category) of the stimulus image, but typically, these properties are not reconstructed together. In this context, we introduce a novel three-stage visual reconstruction approach called the Dual-guided Brain Diffusion Model (DBDM). Initially, we employ the Very Deep Variational Autoencoder (VDVAE) to reconstruct a coarse image from fMRI data, capturing the underlying details of the original image. Subsequently, the Bootstrapping Language-Image Pre-training (BLIP) model is utilized to provide a semantic annotation for each image. Finally, the image-to-image generation pipeline of the Versatile Diffusion (VD) model is utilized to recover natural images from the fMRI patterns guided by both visual and semantic information. The experimental results demonstrate that DBDM surpasses previous approaches in both qualitative and quantitative comparisons. In particular, the best performance is achieved by DBDM in reconstructing the semantic details of the original image; the Inception, CLIP and SwAV distances are 0.611, 0.225 and 0.405, respectively. This confirms the efficacy of our model and its potential to advance visual decoding research.

Natural scene reconstruction from fMRI signals using generative latent diffusion

In neural decoding research, one of the most intriguing topics is the reconstruction of perceived natural images based on fMRI signals. Previous studies have succeeded in re-creating different aspects of the visuals, such as low-level properties (shape, texture, layout) or high-level features (category of objects, descriptive semantics of scenes) but have typically failed to reconstruct these properties together for complex scene images. Generative AI has recently made a leap forward with latent diffusion models capable of generating high-complexity images. Here, we investigate how to take advantage of this innovative technology for brain decoding. We present a two-stage scene reconstruction framework called "Brain-Diffuser". In the first stage, starting from fMRI signals, we reconstruct images that capture low-level properties and overall layout using a VDVAE (Very Deep Variational Autoencoder) model. In the second stage, we use the image-to-image framework of a latent diffusion model (Versatile Diffusion) conditioned on predicted multimodal (text and visual) features, to generate final reconstructed images. On the publicly available Natural Scenes Dataset benchmark, our method outperforms previous models both qualitatively and quantitatively. When applied to synthetic fMRI patterns generated from individual ROI (region-of-interest) masks, our trained model creates compelling "ROI-optimal" scenes consistent with neuroscientific knowledge. Thus, the proposed methodology can have an impact on both applied (e.g. brain-computer interface) and fundamental neuroscience.

Decoding Visual Neural Representations by Multimodal Learning of Brain-Visual-Linguistic Features

Decoding human visual neural representations is a challenging task with great scientific significance in revealing vision-processing mechanisms and developing brain-like intelligent machines. Most existing methods are difficult to generalize to novel categories that have no corresponding neural data for training. The two main reasons are 1) the under-exploitation of the multimodal semantic knowledge underlying the neural data and 2) the small number of paired (stimuli-responses) training data. To overcome these limitations, this paper presents a generic neural decoding method called BraVL that uses multimodal learning of brain-visual-linguistic features. We focus on modeling the relationships between brain, visual and linguistic features via multimodal deep generative models. Specifically, we leverage the mixture-of-product-of-experts formulation to infer a latent code that enables a coherent joint generation of all three modalities. To learn a more consistent joint representation and improve the data efficiency in the case of limited brain activity data, we exploit both intra- and inter-modality mutual information maximization regularization terms. In particular, our BraVL model can be trained under various semi-supervised scenarios to incorporate the visual and textual features obtained from the extra categories. Finally, we construct three trimodal matching datasets, and the extensive experiments lead to some interesting conclusions and cognitive insights: 1) decoding novel visual categories from human brain activity is practically possible with good accuracy; 2) decoding models using the combination of visual and linguistic features perform much better than those using either of them alone; 3) visual perception may be accompanied by linguistic influences to represent the semantics of visual stimuli.

Reconstructing controllable faces from brain activity with hierarchical multiview representations

Reconstructing visual experience from brain responses measured by functional magnetic resonance imaging (fMRI) is a challenging yet important research topic in brain decoding, especially it has proved more difficult to decode visually similar stimuli, such as faces. Although face attributes are known as the key to face recognition, most existing methods generally ignore how to decode facial attributes more precisely in perceived face reconstruction, which often leads to indistinguishable reconstructed faces. To solve this problem, we propose a novel neural decoding framework called VSPnet (voxel2style2pixel) by establishing hierarchical encoding and decoding networks with disentangled latent representations as media, so that to recover visual stimuli more elaborately. And we design a hierarchical visual encoder (named HVE) to pre-extract features containing both high-level semantic knowledge and low-level visual details from stimuli. The proposed VSPnet consists of two networks: Multi-branch cognitive encoder and style-based image generator. The encoder network is constructed by multiple linear regression branches to map brain signals to the latent space provided by the pre-extracted visual features and obtain representations containing hierarchical information consistent to the corresponding stimuli. We make the generator network inspired by StyleGAN to untangle the complexity of fMRI representations and generate images. And the HVE network is composed of a standard feature pyramid over a ResNet backbone. Extensive experimental results on the latest public datasets have demonstrated the reconstruction accuracy of our proposed method outperforms the state-of-the-art approaches and the identifiability of different reconstructed faces has been greatly improved. In particular, we achieve feature editing for several facial attributes in fMRI domain based on the multiview (i.e., visual stimuli and evoked fMRI) latent representations.

EEG-based BCI Dataset of Semantic Concepts for Imagination and Perception Tasks

Electroencephalography (EEG) is a widely-used neuroimaging technique in Brain Computer Interfaces (BCIs) due to its non-invasive nature, accessibility and high temporal resolution. A range of input representations has been explored for BCIs. The same semantic meaning can be conveyed in different representations, such as visual (orthographic and pictorial) and auditory (spoken words). These stimuli representations can be either imagined or perceived by the BCI user. In particular, there is a scarcity of existing open source EEG datasets for imagined visual content, and to our knowledge there are no open source EEG datasets for semantics captured through multiple sensory modalities for both perceived and imagined content. Here we present an open source multisensory imagination and perception dataset, with twelve participants, acquired with a 124 EEG channel system. The aim is for the dataset to be open for purposes such as BCI related decoding and for better understanding the neural mechanisms behind perception, imagination and across the sensory modalities when the semantic category is held constant.

Bimodal electroencephalography-functional magnetic resonance imaging dataset for inner-speech recognition

The recognition of inner speech, which could give a 'voice' to patients that have no ability to speak or move, is a challenge for brain-computer interfaces (BCIs). A shortcoming of the available datasets is that they do not combine modalities to increase the performance of inner speech recognition. Multimodal datasets of brain data enable the fusion of neuroimaging modalities with complimentary properties, such as the high spatial resolution of functional magnetic resonance imaging (fMRI) and the temporal resolution of electroencephalography (EEG), and therefore are promising for decoding inner speech. This paper presents the first publicly available bimodal dataset containing EEG and fMRI data acquired nonsimultaneously during inner-speech production. Data were obtained from four healthy, right-handed participants during an inner-speech task with words in either a social or numerical category. Each of the 8-word stimuli were assessed with 40 trials, resulting in 320 trials in each modality for each participant. The aim of this work is to provide a publicly available bimodal dataset on inner speech, contributing towards speech prostheses.

Inter-individual deep image reconstruction via hierarchical neural code conversion

The sensory cortex is characterized by general organizational principles such as topography and hierarchy. However, measured brain activity given identical input exhibits substantially different patterns across individuals. Although anatomical and functional alignment methods have been proposed in functional magnetic resonance imaging (fMRI) studies, it remains unclear whether and how hierarchical and fine-grained representations can be converted between individuals while preserving the encoded perceptual content. In this study, we trained a method of functional alignment called neural code converter that predicts a target subject's brain activity pattern from a source subject given the same stimulus, and analyzed the converted patterns by decoding hierarchical visual features and reconstructing perceived images. The converters were trained on fMRI responses to identical sets of natural images presented to pairs of individuals, using the voxels on the visual cortex that covers from V1 through the ventral object areas without explicit labels of the visual areas. We decoded the converted brain activity patterns into the hierarchical visual features of a deep neural network using decoders pre-trained on the target subject and then reconstructed images via the decoded features. Without explicit information about the visual cortical hierarchy, the converters automatically learned the correspondence between visual areas of the same levels. Deep neural network feature decoding at each layer showed higher decoding accuracies from corresponding levels of visual areas, indicating that hierarchical representations were preserved after conversion. The visual images were reconstructed with recognizable silhouettes of objects even with relatively small numbers of data for converter training. The decoders trained on pooled data from multiple individuals through conversions led to a slight improvement over those trained on a single individual. These results demonstrate that the hierarchical and fine-grained representation can be converted by functional alignment, while preserving sufficient visual information to enable inter-individual visual image reconstruction.

Neurophysiology, Neuropsychology, and Epilepsy, in 2022: Hills We Have Climbed and Hills Ahead. Neurophysiology in epilepsy

Since the discovery of the human electroencephalogram (EEG), neurophysiology techniques have become indispensable tools in our armamentarium to localize epileptic seizures. New signal analysis techniques and the prospects of artificial intelligence and big data will offer unprecedented opportunities to further advance the field in the near future, ultimately resulting in improved quality of life for many patients with drug-resistant epilepsy. This article summarizes selected presentations from Day 1 of the two-day symposium "Neurophysiology, Neuropsychology, Epilepsy, 2022: Hills We Have Climbed and the Hills Ahead". Day 1 was dedicated to highlighting and honoring the work of Dr. Jean Gotman, a pioneer in EEG, intracranial EEG, simultaneous EEG/ functional magnetic resonance imaging, and signal analysis of epilepsy. The program focused on two main research directions of Dr. Gotman, and was dedicated to "High-frequency oscillations, a new biomarker of epilepsy" and "Probing the epileptic focus from inside and outside". All talks were presented by colleagues and former trainees of Dr. Gotman. The extended summaries provide an overview of historical and current work in the neurophysiology of epilepsy with emphasis on novel EEG biomarkers of epilepsy and source imaging and concluded with an outlook on the future of epilepsy research, and what is needed to bring the field to the next level.

Diverse Task Classification from Activation Patterns of Functional Neuro-Images Using Feature Fusion Module

One of the emerging fields in functional magnetic resonance imaging (fMRI) is the decoding of different stimulations. The underlying idea is to reveal the hidden representative signal patterns of various fMRI tasks for achieving high task-classification performance. Unfortunately, when multiple tasks are processed, performance remains limited due to several challenges, which are rarely addressed since the majority of the state-of-the-art studies cover a single neuronal activity task. Accordingly, the first contribution of this study is the collection and release of a rigorously acquired dataset, which contains cognitive, behavioral, and affective fMRI tasks together with resting state. After a comprehensive analysis of the pitfalls of existing systems on this new dataset, we propose an automatic multitask classification (MTC) strategy using a feature fusion module (FFM). FFM aims to create a unique signature for each task by combining deep features with time-frequency representations. We show that FFM creates a feature space that is superior for representing task characteristics compared to their individual use. Finally, for MTC, we test a diverse set of deep-models and analyze their complementarity. Our results reveal higher classification accuracy compared to benchmarks. Both the dataset and the code are accessible to researchers for further developments.

Neural representations of the perception of handwritten digits and visual objects from a convolutional neural network compared to humans

We investigated neural representations for visual perception of 10 handwritten digits and six visual objects from a convolutional neural network (CNN) and humans using functional magnetic resonance imaging (fMRI). Once our CNN model was fine-tuned using a pre-trained VGG16 model to recognize the visual stimuli from the digit and object categories, representational similarity analysis (RSA) was conducted using neural activations from fMRI and feature representations from the CNN model across all 16 classes. The encoded neural representation of the CNN model exhibited the hierarchical topography mapping of the human visual system. The feature representations in the lower convolutional (Conv) layers showed greater similarity with the neural representations in the early visual areas and parietal cortices, including the posterior cingulate cortex. The feature representations in the higher Conv layers were encoded in the higher-order visual areas, including the ventral/medial/dorsal stream and middle temporal complex. The neural representations in the classification layers were observed mainly in the ventral stream visual cortex (including the inferior temporal cortex), superior parietal cortex, and prefrontal cortex. There was a surprising similarity between the neural representations from the CNN model and the neural representations for human visual perception in the context of the perception of digits versus objects, particularly in the primary visual and associated areas. This study also illustrates the uniqueness of human visual perception. Unlike the CNN model, the neural representation of digits and objects for humans is more widely distributed across the whole brain, including the frontal and temporal areas.

Brain2Pix: Fully convolutional naturalistic video frame reconstruction from brain activity

Reconstructing complex and dynamic visual perception from brain activity remains a major challenge in machine learning applications to neuroscience. Here, we present a new method for reconstructing naturalistic images and videos from very large single-participant functional magnetic resonance imaging data that leverages the recent success of image-to-image transformation networks. This is achieved by exploiting spatial information obtained from retinotopic mappings across the visual system. More specifically, we first determine what position each voxel in a particular region of interest would represent in the visual field based on its corresponding receptive field location. Then, the 2D image representation of the brain activity on the visual field is passed to a fully convolutional image-to-image network trained to recover the original stimuli using VGG feature loss with an adversarial regularizer. In our experiments, we show that our method offers a significant improvement over existing video reconstruction techniques.

Neural Decoding of Visual Information Across Different Neural Recording Modalities and Approaches

Vision plays a peculiar role in intelligence. Visual information, forming a large part of the sensory information, is fed into the human brain to formulate various types of cognition and behaviours that make humans become intelligent agents. Recent advances have led to the development of brain-inspired algorithms and models for machine vision. One of the key components of these methods is the utilization of the computational principles underlying biological neurons. Additionally, advanced experimental neuroscience techniques have generated different types of neural signals that carry essential visual information. Thus, there is a high demand for mapping out functional models for reading out visual information from neural signals. Here, we briefly review recent progress on this issue with a focus on how machine learning techniques can help in the development of models for contending various types of neural signals, from fine-scale neural spikes and single-cell calcium imaging to coarse-scale electroencephalography (EEG) and functional magnetic resonance imaging recordings of brain signals.

Brain Decoding of Multiple Subjects for Estimating Visual Information Based on a Probabilistic Generative Model

Brain decoding is a process of decoding human cognitive contents from brain activities. However, improving the accuracy of brain decoding remains difficult due to the unique characteristics of the brain, such as the small sample size and high dimensionality of brain activities. Therefore, this paper proposes a method that effectively uses multi-subject brain activities to improve brain decoding accuracy. Specifically, we distinguish between the shared information common to multi-subject brain activities and the individual information based on each subject's brain activities, and both types of information are used to decode human visual cognition. Both types of information are extracted as features belonging to a latent space using a probabilistic generative model. In the experiment, an publicly available dataset and five subjects were used, and the estimation accuracy was validated on the basis of a confidence score ranging from 0 to 1, and a large value indicates superiority. The proposed method achieved a confidence score of 0.867 for the best subject and an average of 0.813 for the five subjects, which was the best compared to other methods. The experimental results show that the proposed method can accurately decode visual cognition compared with other existing methods in which the shared information is not distinguished from the individual information.

Self-supervised Natural Image Reconstruction and Large-scale Semantic Classification from Brain Activity

Reconstructing natural images and decoding their semantic category from fMRI brain recordings is challenging. Acquiring sufficient pairs of images and their corresponding fMRI responses, which span the huge space of natural images, is prohibitive. We present a novel self-supervised approach that goes well beyond the scarce paired data, for achieving both: (i) state-of-the art fMRI-to-image reconstruction, and (ii) first-ever large-scale semantic classification from fMRI responses. By imposing cycle consistency between a pair of deep neural networks (from image-to-fMRI & from fMRI-to-image), we train our image reconstruction network on a large number of "unpaired" natural images (images without fMRI recordings) from many novel semantic categories. This enables to adapt our reconstruction network to a very rich semantic coverage without requiring any explicit semantic supervision. Specifically, we find that combining our self-supervised training with high-level perceptual losses, gives rise to new reconstruction & classification capabilities. In particular, this perceptual training enables to classify well fMRIs of never-before-seen semantic classes, without requiring any class labels during training. This gives rise to: (i) Unprecedented image-reconstruction from fMRI of never-before-seen images (evaluated by image metrics and human testing), and (ii) Large-scale semantic classification of categories that were never-before-seen during network training. Such large-scale (1000-way) semantic classification from fMRI recordings has never been demonstrated before. Finally, we provide evidence for the biological consistency of our learned model.

Face identity coding in the deep neural network and primate brain

A central challenge in face perception research is to understand how neurons encode face identities. This challenge has not been met largely due to the lack of simultaneous access to the entire face processing neural network and the lack of a comprehensive multifaceted model capable of characterizing a large number of facial features. Here, we addressed this challenge by conducting in silico experiments using a pre-trained face recognition deep neural network (DNN) with a diverse array of stimuli. We identified a subset of DNN units selective to face identities, and these identity-selective units demonstrated generalized discriminability to novel faces. Visualization and manipulation of the network revealed the importance of identity-selective units in face recognition. Importantly, using our monkey and human single-neuron recordings, we directly compared the response of artificial units with real primate neurons to the same stimuli and found that artificial units shared a similar representation of facial features as primate neurons. We also observed a region-based feature coding mechanism in DNN units as in human neurons. Together, by directly linking between artificial and primate neural systems, our results shed light on how the primate brain performs face recognition tasks.

Deep Residual Convolutional Neural Networks for Brain-Computer Interface to Visualize Neural Processing of Hand Movements in the Human Brain

Concomitant with the development of deep learning, brain-computer interface (BCI) decoding technology has been rapidly evolving. Convolutional neural networks (CNNs), which are generally used as electroencephalography (EEG) classification models, are often deployed in BCI prototypes to improve the estimation accuracy of a participant's brain activity. However, because most BCI models are trained, validated, and tested via within-subject cross-validation and there is no corresponding generalization model, their applicability to unknown participants is not guaranteed. In this study, to facilitate the generalization of BCI model performance to unknown participants, we trained a model comprising multiple layers of residual CNNs and visualized the reasons for BCI classification to reveal the location and timing of neural activities that contribute to classification. Specifically, to develop a BCI that can distinguish between rest, left-hand movement, and right-hand movement tasks with high accuracy, we created multilayers of CNNs, inserted residual networks into the multilayers, and used a larger dataset than in previous studies. The constructed model was analyzed with gradient-class activation mapping (Grad-CAM). We evaluated the developed model via subject cross-validation and found that it achieved significantly improved accuracy (85.69 ± 1.10%) compared with conventional models or without residual networks. Grad-CAM analysis of the classification of cases in which our model produced correct answers showed localized activity near the premotor cortex. These results confirm the effectiveness of inserting residual networks into CNNs for tuning BCI. Further, they suggest that recording EEG signals over the premotor cortex and some other areas contributes to high classification accuracy.

A studyforrest extension, MEG recordings while watching the audio-visual movie "Forrest Gump"

Naturalistic stimuli, such as movies, are being increasingly used to map brain function because of their high ecological validity. The pioneering studyforrest and other naturalistic neuroimaging projects have provided free access to multiple movie-watching functional magnetic resonance imaging (fMRI) datasets to prompt the community for naturalistic experimental paradigms. However, sluggish blood-oxygenation-level-dependent fMRI signals are incapable of resolving neuronal activity with the temporal resolution at which it unfolds. Instead, magnetoencephalography (MEG) measures changes in the magnetic field produced by neuronal activity and is able to capture rich dynamics of the brain at the millisecond level while watching naturalistic movies. Herein, we present the first public prolonged MEG dataset collected from 11 participants while watching the 2 h long audio-visual movie "Forrest Gump". Minimally preprocessed data was also provided to facilitate the use of the dataset. As a studyforrest extension, we envision that this dataset, together with fMRI data from the studyforrest project, will serve as a foundation for exploring the neural dynamics of various cognitive functions in real-world contexts.

Voluntary control of semantic neural representations by imagery with conflicting visual stimulation

Neural representations of visual perception are affected by mental imagery and attention. Although attention is known to modulate neural representations, it is unknown how imagery changes neural representations when imagined and perceived images semantically conflict. We hypothesized that imagining an image would activate a neural representation during its perception even while watching a conflicting image. To test this hypothesis, we developed a closed-loop system to show images inferred from electrocorticograms using a visual semantic space. The successful control of the feedback images demonstrated that the semantic vector inferred from electrocorticograms became closer to the vector of the imagined category, even while watching images from different categories. Moreover, modulation of the inferred vectors by mental imagery depended asymmetrically on the perceived and imagined categories. Shared neural representation between mental imagery and perception was still activated by the imagery under semantically conflicting perceptions depending on the semantic category.

In this study, intracranial EEG recordings show that neural representations of imagined images can still be present in humans even when they are shown conflicting images.

NeuroGen: Activation optimized image synthesis for discovery neuroscience

Functional MRI (fMRI) is a powerful technique that has allowed us to characterize visual cortex responses to stimuli, yet such experiments are by nature constructed based on a priori hypotheses, limited to the set of images presented to the individual while they are in the scanner, are subject to noise in the observed brain responses, and may vary widely across individuals. In this work, we propose a novel computational strategy, which we call NeuroGen, to overcome these limitations and develop a powerful tool for human vision neuroscience discovery. NeuroGen combines an fMRI-trained neural encoding model of human vision with a deep generative network to synthesize images predicted to achieve a target pattern of macro-scale brain activation. We demonstrate that the reduction of noise that the encoding model provides, coupled with the generative network's ability to produce images of high fidelity, results in a robust discovery architecture for visual neuroscience. By using only a small number of synthetic images created by NeuroGen, we demonstrate that we can detect and amplify differences in regional and individual human brain response patterns to visual stimuli. We then verify that these discoveries are reflected in the several thousand observed image responses measured with fMRI. We further demonstrate that NeuroGen can create synthetic images predicted to achieve regional response patterns not achievable by the best-matching natural images. The NeuroGen framework extends the utility of brain encoding models and opens up a new avenue for exploring, and possibly precisely controlling, the human visual system.

fMRI Brain Decoding and Its Applications in Brain-Computer Interface: A Survey

Brain neural activity decoding is an important branch of neuroscience research and a key technology for the brain-computer interface (BCI). Researchers initially developed simple linear models and machine learning algorithms to classify and recognize brain activities. With the great success of deep learning on image recognition and generation, deep neural networks (DNN) have been engaged in reconstructing visual stimuli from human brain activity via functional magnetic resonance imaging (fMRI). In this paper, we reviewed the brain activity decoding models based on machine learning and deep learning algorithms. Specifically, we focused on current brain activity decoding models with high attention: variational auto-encoder (VAE), generative confrontation network (GAN), and the graph convolutional network (GCN). Furthermore, brain neural-activity-decoding-enabled fMRI-based BCI applications in mental and psychological disease treatment are presented to illustrate the positive correlation between brain decoding and BCI. Finally, existing challenges and future research directions are addressed.

A CNN-transformer hybrid approach for decoding visual neural activity into text

BACKGROUND AND OBJECTIVE

Most studies used neural activities evoked by linguistic stimuli such as phrases or sentences to decode the language structure. However, compared to linguistic stimuli, it is more common for the human brain to perceive the outside world through non-linguistic stimuli such as natural images, so only relying on linguistic stimuli cannot fully understand the information perceived by the human brain. To address this, an end-to-end mapping model between visual neural activities evoked by non-linguistic stimuli and visual contents is demanded.

METHODS

Inspired by the success of the Transformer network in neural machine translation and the convolutional neural network (CNN) in computer vision, here a CNN-Transformer hybrid language decoding model is constructed in an end-to-end fashion to decode functional magnetic resonance imaging (fMRI) signals evoked by natural images into descriptive texts about the visual stimuli. Specifically, this model first encodes a semantic sequence extracted by a two-layer 1D CNN from the multi-time visual neural activity into a multi-level abstract representation, then decodes this representation, step by step, into an English sentence.

RESULTS

Experimental results show that the decoded texts are semantically consistent with the corresponding ground truth annotations. Additionally, by varying the encoding and decoding layers and modifying the original positional encoding of the Transformer, we found that a specific architecture of the Transformer is required in this work.

CONCLUSIONS

The study results indicate that the proposed model can decode the visual neural activities evoked by natural images into descriptive text about the visual stimuli in the form of sentences. Hence, it may be considered as a potential computer-aided tool for neuroscientists to understand the neural mechanism of visual information processing in the human brain in the future.

Structured Neural Decoding With Multitask Transfer Learning of Deep Neural Network Representations

The reconstruction of visual information from human brain activity is a very important research topic in brain decoding. Existing methods ignore the structural information underlying the brain activities and the visual features, which severely limits their performance and interpretability. Here, we propose a hierarchically structured neural decoding framework by using multitask transfer learning of deep neural network (DNN) representations and a matrix-variate Gaussian prior. Our framework consists of two stages, Voxel2Unit and Unit2Pixel. In Voxel2Unit, we decode the functional magnetic resonance imaging (fMRI) data to the intermediate features of a pretrained convolutional neural network (CNN). In Unit2Pixel, we further invert the predicted CNN features back to the visual images. Matrix-variate Gaussian prior allows us to take into account the structures between feature dimensions and between regression tasks, which are useful for improving decoding effectiveness and interpretability. This is in contrast with the existing single-output regression models that usually ignore these structures. We conduct extensive experiments on two real-world fMRI data sets, and the results show that our method can predict CNN features more accurately and reconstruct the perceived natural images and faces with higher quality.

Attention modulates neural representation to render reconstructions according to subjective appearance

Stimulus images can be reconstructed from visual cortical activity. However, our perception of stimuli is shaped by both stimulus-induced and top-down processes, and it is unclear whether and how reconstructions reflect top-down aspects of perception. Here, we investigate the effect of attention on reconstructions using fMRI activity measured while subjects attend to one of two superimposed images. A state-of-the-art method is used for image reconstruction, in which brain activity is translated (decoded) to deep neural network (DNN) features of hierarchical layers then to an image. Reconstructions resemble the attended rather than unattended images. They can be modeled by superimposed images with biased contrasts, comparable to the appearance during attention. Attentional modulations are found in a broad range of hierarchical visual representations and mirror the brain-DNN correspondence. Our results demonstrate that top-down attention counters stimulus-induced responses, modulating neural representations to render reconstructions in accordance with subjective appearance.

Hyperrealistic neural decoding for reconstructing faces from fMRI activations via the GAN latent space

Neural decoding can be conceptualized as the problem of mapping brain responses back to sensory stimuli via a feature space. We introduce (i) a novel experimental paradigm that uses well-controlled yet highly naturalistic stimuli with a priori known feature representations and (ii) an implementation thereof for HYPerrealistic reconstruction of PERception (HYPER) of faces from brain recordings. To this end, we embrace the use of generative adversarial networks (GANs) at the earliest step of our neural decoding pipeline by acquiring fMRI data as participants perceive face images synthesized by the generator network of a GAN. We show that the latent vectors used for generation effectively capture the same defining stimulus properties as the fMRI measurements. As such, these latents (conditioned on the GAN) are used as the in-between feature representations underlying the perceived images that can be predicted in neural decoding for (re-)generation of the originally perceived stimuli, leading to the most accurate reconstructions of perception to date.

Natural Image Reconstruction From fMRI Using Deep Learning: A Survey

With the advent of brain imaging techniques and machine learning tools, much effort has been devoted to building computational models to capture the encoding of visual information in the human brain. One of the most challenging brain decoding tasks is the accurate reconstruction of the perceived natural images from brain activities measured by functional magnetic resonance imaging (fMRI). In this work, we survey the most recent deep learning methods for natural image reconstruction from fMRI. We examine these methods in terms of architectural design, benchmark datasets, and evaluation metrics and present a fair performance evaluation across standardized evaluation metrics. Finally, we discuss the strengths and limitations of existing studies and present potential future directions.

Photorealistic Reconstruction of Visual Texture From EEG Signals

Recent advances in brain decoding have made it possible to classify image categories based on neural activity. Increasing numbers of studies have further attempted to reconstruct the image itself. However, because images of objects and scenes inherently involve spatial layout information, the reconstruction usually requires retinotopically organized neural data with high spatial resolution, such as fMRI signals. In contrast, spatial layout does not matter in the perception of "texture," which is known to be represented as spatially global image statistics in the visual cortex. This property of "texture" enables us to reconstruct the perceived image from EEG signals, which have a low spatial resolution. Here, we propose an MVAE-based approach for reconstructing texture images from visual evoked potentials measured from observers viewing natural textures such as the textures of various surfaces and object ensembles. This approach allowed us to reconstruct images that perceptually resemble the original textures with a photographic appearance. The present approach can be used as a method for decoding the highly detailed "impression" of sensory stimuli from brain activity.

Natural and Artificial Intelligence: A brief introduction to the interplay between AI and neuroscience research

Neuroscience and artificial intelligence (AI) share a long history of collaboration. Advances in neuroscience, alongside huge leaps in computer processing power over the last few decades, have given rise to a new generation of in silico neural networks inspired by the architecture of the brain. These AI systems are now capable of many of the advanced perceptual and cognitive abilities of biological systems, including object recognition and decision making. Moreover, AI is now increasingly being employed as a tool for neuroscience research and is transforming our understanding of brain functions. In particular, deep learning has been used to model how convolutional layers and recurrent connections in the brain's cerebral cortex control important functions, including visual processing, memory, and motor control. Excitingly, the use of neuroscience-inspired AI also holds great promise for understanding how changes in brain networks result in psychopathologies, and could even be utilized in treatment regimes. Here we discuss recent advancements in four areas in which the relationship between neuroscience and AI has led to major advancements in the field; (1) AI models of working memory, (2) AI visual processing, (3) AI analysis of big neuroscience datasets, and (4) computational psychiatry.

Brain hierarchy score: Which deep neural networks are hierarchically brain-like?

Achievement of human-level image recognition by deep neural networks (DNNs) has spurred interest in whether and how DNNs are brain-like. Both DNNs and the visual cortex perform hierarchical processing, and correspondence has been shown between hierarchical visual areas and DNN layers in representing visual features. Here, we propose the brain hierarchy (BH) score as a metric to quantify the degree of hierarchical correspondence based on neural decoding and encoding analyses where DNN unit activations and human brain activity are predicted from each other. We find that BH scores for 29 pre-trained DNNs with various architectures are negatively correlated with image recognition performance, thus indicating that recently developed high-performance DNNs are not necessarily brain-like. Experimental manipulations of DNN models suggest that single-path sequential feedforward architecture with broad spatial integration is critical to brain-like hierarchy. Our method may provide new ways to design DNNs in light of their representational homology to the brain.

•

A measure for brain-like hierarchy is proposed to characterize DNNs

•

Encoding/decoding with human fMRI quantifies the hierarchical correspondence

•

Among representative DNN models, high-performance models are not brain-like

•

Critical factors for brain-like hierarchy are explored

A neural decoding algorithm that generates language from visual activity evoked by natural images

Transforming neural activities into language is revolutionary for human-computer interaction as well as functional restoration of aphasia. Present rapid development of artificial intelligence makes it feasible to decode the neural signals of human visual activities. In this paper, a novel Progressive Transfer Language Decoding Model (PT-LDM) is proposed to decode visual fMRI signals into phrases or sentences when natural images are being watched. The PT-LDM consists of an image-encoder, a fMRI encoder and a language-decoder. The results showed that phrases and sentences were successfully generated from visual activities. Similarity analysis showed that three often-used evaluation indexes BLEU, ROUGE and CIDEr reached 0.182, 0.197 and 0.680 averagely between the generated texts and the corresponding annotated texts in the testing set respectively, significantly higher than the baseline. Moreover, we found that higher visual areas usually had better performance than lower visual areas and the contribution curve of visual response patterns in language decoding varied at successively different time points. Our findings demonstrate that the neural representations elicited in visual cortices when scenes are being viewed have already contained semantic information that can be utilized to generate human language. Our study shows potential application of language-based brain-machine interfaces in the future, especially for assisting aphasics in communicating more efficiently with fMRI signals.

Convolutional Neural Networks as a Model of the Visual System: Past, Present, and Future

Convolutional neural networks (CNNs) were inspired by early findings in the study of biological vision. They have since become successful tools in computer vision and state-of-the-art models of both neural activity and behavior on visual tasks. This review highlights what, in the context of CNNs, it means to be a good model in computational neuroscience and the various ways models can provide insight. Specifically, it covers the origins of CNNs and the methods by which we validate them as models of biological vision. It then goes on to elaborate on what we can learn about biological vision by understanding and experimenting on CNNs and discusses emerging opportunities for the use of CNNs in vision research beyond basic object recognition.

Generative adversarial networks unlock new methods for cognitive science

Generative adversarial networks (GANs) enable computers to learn complex data distributions and sample from these distributions. When applied to the visual domain, this allows artificial, yet photorealistic images to be synthesized. Their success at this very challenging task triggered an explosion of research within the field of artificial intelligence (AI), yielding various new GAN findings and applications. After explaining the core principles behind GANs and reviewing recent GAN innovations, we illustrate how they can be applied to tackle thorny theoretical and methodological problems in cognitive science. We focus on how GANs can reveal hidden structure in internal representations and how they offer a valuable new compromise in the trade-off between experimental control and ecological validity.