Search

Brain decoding techniques are essential for understanding the neurocognitive system. Although numerous methods have been introduced in this field, accurately aligning complex external stimuli with brain activities remains a formidable challenge. To alleviate alignment difficulties, many studies have simplified their models by employing single-task paradigms and establishing direct links between brain/world through classification strategies. Despite improvements in decoding accuracy, this strategy frequently encounters issues with generality when adapting these models to various task paradigms. To address this issue, this study introduces a user-friendly decoding model that enables dynamic communication with the brain, as opposed to the static decoding approaches utilized by traditional studies. The model functions as a brain simulator, allowing for interactive engagement with the brain and enabling the decoding of a subject's experiences through dialogue-like queries. Uniquely, our model is trained in a completely unsupervised and task-free manner. Our experiments demonstrate the feasibility and versatility of our proposed method. Notably, our model demonstrates exceptional capabilities in signal compression, successfully representing the entire brain signal of approximately 185,751 voxels with just 32 signals. Furthermore, we show how our model can integrate seamlessly with multimodal models, thus enhancing the potential for controlling brain decoding through textual or image inputs.

In the medical multi-modal frameworks, the alignment of cross-modality features presents a significant challenge. However, existing works have learned features that are implicitly aligned from the data, without considering the explicit relationships in the medical context. This data-reliance may lead to low generalization of the learned alignment relationships. In this work, we propose the Eye-gaze Guided Multi-modal Alignment (EGMA) framework to harness eye-gaze data for better alignment of medical visual and textual features. We explore the natural auxiliary role of radiologists' eye-gaze data in aligning medical images and text, and introduce a novel approach by using eye-gaze data, collected synchronously by radiologists during diagnostic evaluations. We conduct downstream tasks of image classification and image-text retrieval on four medical datasets, where EGMA achieved state-of-the-art performance and stronger generalization across different datasets. Additionally, we explore the impact of varying amounts of eye-gaze data on model performance, highlighting the feasibility and utility of integrating this auxiliary data into multi-modal alignment framework., Comment: 12 pages, 6 figures

Semi-supervised learning (SSL) seeks to enhance task performance by training on both labeled and unlabeled data. Mainstream SSL image classification methods mostly optimize a loss that additively combines a supervised classification objective with a regularization term derived solely from unlabeled data. This formulation neglects the potential for interaction between labeled and unlabeled images. In this paper, we introduce InterLUDE, a new approach to enhance SSL made of two parts that each benefit from labeled-unlabeled interaction. The first part, embedding fusion, interpolates between labeled and unlabeled embeddings to improve representation learning. The second part is a new loss, grounded in the principle of consistency regularization, that aims to minimize discrepancies in the model's predictions between labeled versus unlabeled inputs. Experiments on standard closed-set SSL benchmarks and a medical SSL task with an uncurated unlabeled set show clear benefits to our approach. On the STL-10 dataset with only 40 labels, InterLUDE achieves 3.2% error rate, while the best previous method reports 14.9%., Comment: Semi-supervised Learning; Vision Transformers

Automated interpretation of ultrasound imaging of the heart (echocardiograms) could improve the detection and treatment of aortic stenosis (AS), a deadly heart disease. However, existing deep learning pipelines for assessing AS from echocardiograms have two key limitations. First, most methods rely on limited 2D cineloops, thereby ignoring widely available Doppler imaging that contains important complementary information about pressure gradients and blood flow abnormalities associated with AS. Second, obtaining labeled data is difficult. There are often far more unlabeled echocardiogram recordings available, but these remain underutilized by existing methods. To overcome these limitations, we introduce Semi-supervised Multimodal Multiple-Instance Learning (SMMIL), a new deep learning framework for automatic interpretation for structural heart diseases like AS. When deployed, SMMIL can combine information from two input modalities, spectral Dopplers and 2D cineloops, to produce a study-level AS diagnosis. During training, SMMIL can combine a smaller labeled set and an abundant unlabeled set of both modalities to improve its classifier. Experiments demonstrate that SMMIL outperforms recent alternatives at 3-level AS severity classification as well as several clinically relevant AS detection tasks., Comment: Echocardiography; Multimodal; Semi-supervised Learning; Multiple-Instance Learning

In this paper, we present a large-scale evaluation probing GPT-4V's capabilities and limitations for biomedical image analysis. GPT-4V represents a breakthrough in artificial general intelligence (AGI) for computer vision, with applications in the biomedical domain. We assess GPT-4V's performance across 16 medical imaging categories, including radiology, oncology, ophthalmology, pathology, and more. Tasks include modality recognition, anatomy localization, disease diagnosis, report generation, and lesion detection. The extensive experiments provide insights into GPT-4V's strengths and weaknesses. Results show GPT-4V's proficiency in modality and anatomy recognition but difficulty with disease diagnosis and localization. GPT-4V excels at diagnostic report generation, indicating strong image captioning skills. While promising for biomedical imaging AI, GPT-4V requires further enhancement and validation before clinical deployment. We emphasize responsible development and testing for trustworthy integration of biomedical AGI. This rigorous evaluation of GPT-4V on diverse medical images advances understanding of multimodal large language models (LLMs) and guides future work toward impactful healthcare applications.

We explore the impact of entropy change in deep learning systems via noise injection at different levels, i.e., the latent space and input image. The series of models that employ our methodology are collectively known as Noisy Neural Networks (NoisyNN), with examples such as NoisyViT and NoisyCNN. Noise is conventionally viewed as a harmful perturbation in various deep learning architectures, such as convolutional neural networks (CNNs) and vision transformers (ViTs), as well as different learning tasks like image classification and transfer learning. However, this work shows noise can be an effective way to change the entropy of the learning system. We demonstrate that specific noise can boost the performance of various deep architectures under certain conditions. We theoretically prove the enhancement gained from positive noise by reducing the task complexity defined by information entropy and experimentally show the significant performance gain in large image datasets, such as the ImageNet. Herein, we use the information entropy to define the complexity of the task. We categorize the noise into two types, positive noise (PN) and harmful noise (HN), based on whether the noise can help reduce the complexity of the task. Extensive experiments of CNNs and ViTs have shown performance improvements by proactively injecting positive noise, where we achieved an unprecedented top 1 accuracy of over 95$\%$ on ImageNet. Both theoretical analysis and empirical evidence have confirmed that the presence of positive noise, can benefit the learning process, while the traditionally perceived harmful noise indeed impairs deep learning models. The different roles of noise offer new explanations for deep models on specific tasks and provide a new paradigm for improving model performance. Moreover, it reminds us that we can influence the performance of learning systems via information entropy change., Comment: Information Entropy, NoisyNN, ViT, CNN

Unsupervised domain adaptation (UDA) aims to learn transferable representation across domains. Recently a few UDA works have successfully applied Transformer-based methods and achieved state-of-the-art (SOTA) results. However, it remains challenging when there exists a large domain gap between the source and target domain. Inspired by humans' exceptional transferability abilities to adapt knowledge from familiar to uncharted domains, we try to apply the universally existing organizational structure in the human functional brain networks, i.e., the core-periphery principle to design the Transformer and improve its UDA performance. In this paper, we propose a novel brain-inspired robust core-periphery constrained transformer (RCCT) for unsupervised domain adaptation, which brings a large margin of performance improvement on various datasets. Specifically, in RCCT, the self-attention operation across image patches is rescheduled by an adaptively learned weighted graph with the Core-Periphery structure (CP graph), where the information communication and exchange between images patches are manipulated and controlled by the connection strength, i.e., edge weight of the learned weighted CP graph. Besides, since the data in domain adaptation tasks can be noisy, to improve the model robustness, we intentionally add perturbations to the patches in the latent space to ensure generating robust learned weighted core-periphery graphs. Extensive evaluations are conducted on several widely tested UDA benchmarks. Our proposed RCCT consistently performs best compared to existing works, including 88.3\% on Office-Home, 95.0\% on Office-31, 90.7\% on VisDA-2017, and 46.0\% on DomainNet., Comment: Core-Periphery, ViT, Unsupervised domain adaptation

With the popularity of deep neural networks (DNNs), model interpretability is becoming a critical concern. Many approaches have been developed to tackle the problem through post-hoc analysis, such as explaining how predictions are made or understanding the meaning of neurons in middle layers. Nevertheless, these methods can only discover the patterns or rules that naturally exist in models. In this work, rather than relying on post-hoc schemes, we proactively instill knowledge to alter the representation of human-understandable concepts in hidden layers. Specifically, we use a hierarchical tree of semantic concepts to store the knowledge, which is leveraged to regularize the representations of image data instances while training deep models. The axes of the latent space are aligned with the semantic concepts, where the hierarchical relations between concepts are also preserved. Experiments on real-world image datasets show that our method improves model interpretability, showing better disentanglement of semantic concepts, without negatively affecting model classification performance.

In this study, we evaluate the performance of the Segment Anything Model (SAM) in clinical radiotherapy. Our results indicate that SAM's 'segment anything' mode can achieve clinically acceptable segmentation results in most organs-at-risk (OARs) with Dice scores higher than 0.7. SAM's 'box prompt' mode further improves the Dice scores by 0.1 to 0.5. Considering the size of the organ and the clarity of its boundary, SAM displays better performance for large organs with clear boundaries but performs worse for smaller organs with unclear boundaries. Given that SAM, a model pre-trained purely on natural images, can handle the delineation of OARs from medical images with clinically acceptable accuracy, these results highlight SAM's robust generalization capabilities with consistent accuracy in automatic segmentation for radiotherapy. In other words, SAM can achieve delineation of different OARs at different sites using a generic automatic segmentation model. SAM's generalization capabilities across different disease sites suggest that it is technically feasible to develop a generic model for automatic segmentation in radiotherapy.

Prompts have been proven to play a crucial role in large language models, and in recent years, vision models have also been using prompts to improve scalability for multiple downstream tasks. In this paper, we focus on adapting prompt design based on instruction tuning into a visual transformer model for image classification which we called Instruction-ViT. The key idea is to implement multi-modal prompts (text or image prompt) related to category information to guide the fine-tuning of the model. Based on the experiments of several image captionining tasks, the performance and domain adaptability were improved. Our work provided an innovative strategy to fuse multi-modal prompts with better performance and faster adaptability for visual classification models.

Recently, ChatGPT and GPT-4 have emerged and gained immense global attention due to their unparalleled performance in language processing. Despite demonstrating impressive capability in various open-domain tasks, their adequacy in highly specific fields like radiology remains untested. Radiology presents unique linguistic phenomena distinct from open-domain data due to its specificity and complexity. Assessing the performance of large language models (LLMs) in such specific domains is crucial not only for a thorough evaluation of their overall performance but also for providing valuable insights into future model design directions: whether model design should be generic or domain-specific. To this end, in this study, we evaluate the performance of ChatGPT/GPT-4 on a radiology NLI task and compare it to other models fine-tuned specifically on task-related data samples. We also conduct a comprehensive investigation on ChatGPT/GPT-4's reasoning ability by introducing varying levels of inference difficulty. Our results show that 1) GPT-4 outperforms ChatGPT in the radiology NLI task; 2) other specifically fine-tuned models require significant amounts of data samples to achieve comparable performance to ChatGPT/GPT-4. These findings demonstrate that constructing a generic model that is capable of solving various tasks across different domains is feasible.

Artificial General Intelligence (AGI) has been a long-standing goal of humanity, with the aim of creating machines capable of performing any intellectual task that humans can do. To achieve this, AGI researchers draw inspiration from the human brain and seek to replicate its principles in intelligent machines. Brain-inspired artificial intelligence is a field that has emerged from this endeavor, combining insights from neuroscience, psychology, and computer science to develop more efficient and powerful AI systems. In this article, we provide a comprehensive overview of brain-inspired AI from the perspective of AGI. We begin with the current progress in brain-inspired AI and its extensive connection with AGI. We then cover the important characteristics for both human intelligence and AGI (e.g., scaling, multimodality, and reasoning). We discuss important technologies toward achieving AGI in current AI systems, such as in-context learning and prompt tuning. We also investigate the evolution of AGI systems from both algorithmic and infrastructural perspectives. Finally, we explore the limitations and future of AGI.

Designing more efficient, reliable, and explainable neural network architectures is critical to studies that are based on artificial intelligence (AI) techniques. Previous studies, by post-hoc analysis, have found that the best-performing ANNs surprisingly resemble biological neural networks (BNN), which indicates that ANNs and BNNs may share some common principles to achieve optimal performance in either machine learning or cognitive/behavior tasks. Inspired by this phenomenon, we proactively instill organizational principles of BNNs to guide the redesign of ANNs. We leverage the Core-Periphery (CP) organization, which is widely found in human brain networks, to guide the information communication mechanism in the self-attention of vision transformer (ViT) and name this novel framework as CP-ViT. In CP-ViT, the attention operation between nodes is defined by a sparse graph with a Core-Periphery structure (CP graph), where the core nodes are redesigned and reorganized to play an integrative role and serve as a center for other periphery nodes to exchange information. We evaluated the proposed CP-ViT on multiple public datasets, including medical image datasets (INbreast) and natural image datasets. Interestingly, by incorporating the BNN-derived principle (CP structure) into the redesign of ViT, our CP-ViT outperforms other state-of-the-art ANNs. In general, our work advances the state of the art in three aspects: 1) This work provides novel insights for brain-inspired AI: we can utilize the principles found in BNNs to guide and improve our ANN architecture design; 2) We show that there exist sweet spots of CP graphs that lead to CP-ViTs with significantly improved performance; and 3) The core nodes in CP-ViT correspond to task-related meaningful and important image patches, which can significantly enhance the interpretability of the trained deep model., Comment: Core-periphery, functional brain networks, ViT

The digitization of healthcare has facilitated the sharing and re-using of medical data but has also raised concerns about confidentiality and privacy. HIPAA (Health Insurance Portability and Accountability Act) mandates removing re-identifying information before the dissemination of medical records. Thus, effective and efficient solutions for de-identifying medical data, especially those in free-text forms, are highly needed. While various computer-assisted de-identification methods, including both rule-based and learning-based, have been developed and used in prior practice, such solutions still lack generalizability or need to be fine-tuned according to different scenarios, significantly imposing restrictions in wider use. The advancement of large language models (LLM), such as ChatGPT and GPT-4, have shown great potential in processing text data in the medical domain with zero-shot in-context learning, especially in the task of privacy protection, as these models can identify confidential information by their powerful named entity recognition (NER) capability. In this work, we developed a novel GPT4-enabled de-identification framework (``DeID-GPT") to automatically identify and remove the identifying information. Compared to existing commonly used medical text data de-identification methods, our developed DeID-GPT showed the highest accuracy and remarkable reliability in masking private information from the unstructured medical text while preserving the original structure and meaning of the text. This study is one of the earliest to utilize ChatGPT and GPT-4 for medical text data processing and de-identification, which provides insights for further research and solution development on the use of LLMs such as ChatGPT/GPT-4 in healthcare. Codes and benchmarking data information are available at https://github.com/yhydhx/ChatGPT-API.

The human cerebral cortex is highly convoluted into convex gyri and concave sulci. It has been demonstrated that gyri and sulci are significantly different in their anatomy, connectivity, and function, besides exhibiting opposite shape patterns, long-distance axonal fibers connected to gyri are much denser than those connected to sulci, and neural signals on gyri are more complex in low-frequency while sulci are more complex in high-frequency. Although accumulating evidence shows significant differences between gyri and sulci, their primary roles in brain function have not been elucidated yet. To solve this fundamental problem, we design a novel Twin-Transformer framework to unveil the unique functional roles of gyri and sulci as well as their relationship in the whole brain function. Our Twin-Transformer framework adopts two structure-identical (twin) Transformers to disentangle spatial-temporal patterns of gyri and sulci, one focuses on the information of gyri and the other is on sulci. The Gyro-Sulcal interactions, along with the tremendous but widely existing variability across subjects, are characterized in the loss design. We validated our Twin-Transformer on the HCP task-fMRI dataset, for the first time, to elucidate the different roles of gyri and sulci in brain function. Our results suggest that gyri and sulci could work together in a core-periphery network manner, that is, gyri could serve as core networks for information gathering and distributing, while sulci could serve as periphery networks for specific local information processing. These findings have shed new light on our fundamental understanding of the brain's basic structural and functional mechanisms., Comment: 13 pages, 4 figures

Effective representation of brain anatomical architecture is fundamental in understanding brain regularity and variability. Despite numerous efforts, it is still difficult to infer reliable anatomical correspondence at finer scale, given the tremendous individual variability in cortical folding patterns. It is even more challenging to disentangle common and individual patterns when comparing brains at different neuro-developmental stages. In this work, we developed a novel learning-based few-shot embedding framework to encode the cortical folding patterns into a latent space represented by a group of anatomically meaningful embedding vectors. Specifically, we adopted 3-hinge (3HG) network as the substrate and designed an autoencoder-based embedding framework to learn a common embedding vector for each 3HG's multi-hop feature: each 3HG can be represented as a combination of these feature embeddings via a set of individual specific coefficients to characterize individualized anatomical information. That is, the regularity of folding patterns is encoded into the embeddings, while the individual variations are preserved by the multi=hop combination coefficients. To effectively learn the embeddings for the population with very limited samples, few-shot learning was adopted. We applied our method on adult HCP and pediatric datasets with 1,000+ brains (from 34 gestational weeks to young adult). Our experimental results show that: 1) the learned embedding vectors can quantitatively encode the commonality and individuality of cortical folding patterns; 2) with the embeddings we can robustly infer the complicated many-to-many anatomical correspondences among different brains and 3) our model can be successfully transferred to new populations with very limited training samples.

How to identify and characterize functional brain networks (BN) is fundamental to gain system-level insights into the mechanisms of brain organizational architecture. Current functional magnetic resonance (fMRI) analysis highly relies on prior knowledge of specific patterns in either spatial (e.g., resting-state network) or temporal (e.g., task stimulus) domain. In addition, most approaches aim to find group-wise common functional networks, individual-specific functional networks have been rarely studied. In this work, we propose a novel Twin-Transformers framework to simultaneously infer common and individual functional networks in both spatial and temporal space, in a self-supervised manner. The first transformer takes space-divided information as input and generates spatial features, while the second transformer takes time-related information as input and outputs temporal features. The spatial and temporal features are further separated into common and individual ones via interactions (weights sharing) and constraints between the two transformers. We applied our TwinTransformers to Human Connectome Project (HCP) motor task-fMRI dataset and identified multiple common brain networks, including both task-related and resting-state networks (e.g., default mode network). Interestingly, we also successfully recovered a set of individual-specific networks that are not related to task stimulus and only exist at the individual level., Comment: full pages

An antimagic labeling of a digraph $D$ with $n$ vertices and $m$ arcs is a bijection from the set of arcs of $D$ to $\{1,2,\cdots,m\}$ such that all $n$ oriented vertex-sums are pairwise distinct, where the oriented vertex-sum of a vertex is the sum of labels of all arcs entering that vertex minus the sum of labels of all arcs leaving it. A graph $G$ admits an antimagic orientation if $G$ has an orientation $D$ such that $D$ has an antimagic labeling. Hefetz, M{\"{u}}tze and Schwartz conjectured every connected graph admits an antimagic orientation. In this paper, we support this conjecture by proving that any forest obtained from a given forest with at most one isolated vertex by subdividing each edge at least once admits an antimagic orientation.

Mild cognitive impairment (MCI) is a high-risk dementia condition which progresses to probable Alzheimer's disease (AD) at approximately 10% to 15% per year. Characterization of group-level differences between two subtypes of MCI - stable MCI (sMCI) and progressive MCI (pMCI) is the key step to understand the mechanisms of MCI progression and enable possible delay of transition from MCI to AD. Functional connectivity (FC) is considered as a promising way to study MCI progression since which may show alterations even in preclinical stages and provide substrates for AD progression. However, the representative FC patterns during AD development for different clinical groups, especially for sMCI and pMCI, have been understudied. In this work, we integrated autoencoder and multi-class classification into a single deep model and successfully learned a set of clinical group related feature vectors. Specifically, we trained two non-linear mappings which realized the mutual transformations between original FC space and the feature space. By mapping the learned clinical group related feature vectors to the original FC space, representative FCs were constructed for each group. Moreover, based on these feature vectors, our model achieves a high classification accuracy - 68% for multi-class classification (NC vs SMC vs sMCI vs pMCI vs AD).

A star $k$-edge coloring is a proper edge coloring such that there are no bichromatic paths or cycles of length four. The smallest integer $k$ such that $G$ admits a star $k$-edge coloring is the star chromatic index of $G$. Deng \etal \cite{MR2933839}, and Bezegov{\'a} \etal \cite{MR3431294} independently proved that the star chromatic index of a tree is at most $\lfloor \frac{3\Delta}{2} \rfloor$, and the bound is sharp. Han \etal \cite{MR3924408} strengthened the result to list version of star chromatic index, and proved that $\lfloor \frac{3\Delta}{2} \rfloor$ is also the sharp upper bound for the list star chromatic index of trees. A generalized Halin graph is a plane graph that consists of a plane embedding of a tree $T$ with $\Delta(T) \geq 3$, and a cycle $C$ connecting all the leaves of the tree such that $C$ is the boundary of the exterior face. In this paper, we prove that if $H := T \cup C$ is a generalized Halin graph with $|C| \neq 5$, then its list star chromatic index is at most \[ \max\left\{\left\lfloor\frac{\theta(T) + \Delta(T)}{2}\right\rfloor, 2 \left\lfloor\frac{\Delta(T)}{2}\right\rfloor + 7\right\}, \] where $\theta(T) = \max_{xy \in E(T)}\{d_{T}(x) + d_{T}(y)\}$. As a consequence, if $H$ is a (generalized) Halin graph with maximum degree $\Delta \geq 13$, then the list star chromatic index is at most $\lfloor \frac{3\Delta}{2} \rfloor$. Moreover, the upper bound for the list star chromatic index is sharp.

For a given graph $H$, the Ramsey number $r(H)$ is the minimum $N$ such that any 2-edge-coloring of the complete graph $K_N$ yields a monochromatic copy of $H$. Given a positive integer $n$, a \emph{fan }$F_n$ is a graph formed by $n$ triangles that share one common vertex. We show that ${9n}/{2}-5\le r(F_n)\le {11n}/{2} + 6$ for any $n$. This improves previous best bounds $r(F_n) \le 6n$ of Lin and Li and $r(F_n) \ge 4n+2$ of Zhang, Broersma and Chen.