3.1.1 Stages of Development of Brain Science Research

The number of publications can intuitively measure the developmental trend of the brain science field over the past decade. The peaks and inflection points of the curve can reflect the changes in the research hotspots of the field, which is of great significance for analyzing and predicting the development trend of the brain science field. From the first literature on the topic of brain science research from 1991 until the end of 2012, a period of 22 years, the leading countries in the field of brain science research were the United States, Germany, England, Italy, and France (Table 1). China ranked sixth globally, with 259 papers. However, in the past decade, from 2013 to the end of 2022, the top five countries in terms of publication volume, from high to low, were the United States (2540 papers), China (2103 papers), Germany (1082 papers), the United Kingdom (717 papers), and Canada (528 papers). China's ranking has risen by four, and its total publication volume ranks second. Canada and Australia slightly improved their rankings, while Austria's ranking dropped significantly. The United States, China, Germany, the United Kingdom, and Canada formed the first camp in brain science research in the past decade. Furthermore, China ranked third in brain science research publication volume among the five countries in 2013, with a particularly significant gap with the United States (Figure 2). However, since 2017, China's publication volume has continuously surpassed that of Germany, the United Kingdom, and Italy, and the gap between China and the United States has gradually narrowed. Especially during the COVID-19 pandemic from 2020 to 2022, when the publication volumes of the United States, Germany, the United Kingdom, Italy, and other countries decreased or stagnated to varying degrees, China's publication volume continued to grow rapidly. Since 2020, China has taken the lead and has gradually widened its gap with the second-ranked United States. China has seen the fastest rise in brain science research publication volume in the past decade, especially in the past 5 years, with the most outstanding performance. To explain the reason, thanks to the release of “The 13th Five-Year Plan” by the Chinese government in March 2016, which listed “brain science and brain-like research” as a “major national science and technology innovation and engineering project,” marking the full launch of the “China Brain Project.”

3.1.2 “Low Cohesion, High Separation” in Brain Science Researcher Networks

Co-authorship analysis is helpful in clarifying the current composition of research teams and indicates the maturity of the brain science field, which is a multidimensional reflection of researchers' contributions. Using VOSviewer to create a co-authorship model, a balloon heat map was selected to represent different research teams in different colors, with grayscale indicating continuous exploration of brain science research or smaller amounts of research text. The larger the value of the ball, the higher the researcher's publication volume and contribution to the field. The arcs are used to indicate cooperation between individual nodes, with the thickness of the arcs reflecting the degree of collaboration between the nodes, thus showing the construction and development of a collaborative network structure of a scientific research team.

According to programmed calculations and a two-dimensional diagram, there are 40,897 authors in total, of which 2094 have collaborative relationships, with the minimum value set at 14. Among these, 296 authors formed a collaborative network (Figure 3). Niels Birbaumer, Andrzej Cichocki, Yijun Wang, Jing Jin, and Gao Xiaorong stand out among many researchers, and the overall research team shows a high degree of separation and a low degree of cohesion. Collaboration among researchers is mainly concentrated within their own institutions, and cross-national group cooperation has not yet been formed. There is little exchange and interaction among different groups, and the overall situation is characterized by “separation.” According to the model data, in recent years, there has been a trend in team-oriented construction in brain science research, with Niels Birbaumer, Andrea Kuebler, Gernot R. Mueller-Putz, and Gerwin Schalk as the core team, Andrzej Cichocki and Yu Zhang as the central team, and Yijun Wang, Gao Xiaorong, and Yuanqing Li as the central team. There is indirect cooperation between these teams, whereas the team of Leigh R. Hochberg, Hugues Duffau, and Arthur W. Toga demonstrates only close internal cooperation.

Through statistical analysis of the authors of published articles, we can identify the core authors in the field of brain science research. Based on the number of published articles, the top five authors in the past 20 years are Niels Birbaumer from Eberhard Karls University of Tubingen (with 93 articles), Andrzej Cichocki from Skolkovo Institute of Science and Technology (Skoltech) (with 86 articles), Yijun Wang from the Institute of Semiconductors (with 86 articles), and Jing Jin from Tsinghua University (with 81 articles) (Table 2). Among the highly prolific authors, the number of published papers is generally concentrated between 70 and 90, while the top author has published 93 articles.

3.2.1 High Prominence Country Cooperation

The program data show that the network: N = 105 and E = 881, a total of 105 countries and regions have carried out brain science research over the past decade, with the United States, United Kingdom, France, and Austria forming the primary cooperative group. Cooperation is active, and China has the highest number of publications. However, relatively little international collaboration was observed (Figure 4).

3.2.2 More Centralized Connected Network Clusters

The program data showed that the network had N = 560 and E = 3703. In the past decade, 560 research institutions have conducted brain science research, mainly led by renowned universities. Universities are the main force of global brain science research. From the most critical institutional cooperation map (Figure 5), the Chinese Academy of Sciences in China, the University of California, San Diego in the United States, and University College London in England have the highest prominence in brain science research. The Chinese Academy of Sciences, ranked first, has a centrality of 0.2, followed by the University of California San Diego with a centrality of 0.11 and UCL in England with a centrality of 0.1. However, the four research institutions, Korea University in Korea, the University of Pittsburgh in the United States, Tsinghua University, and Tianjin University in China, have a high publication output and low centrality. These data indicate that their research collaboration with the outside world may still require further strengthening to better integrate into the global brain science research network (Table 3).

3.3.1 Journal Citations Analysis

A total of 192 journals contributed to the publication of brain science research. Over the past decade, the top five journals are “Human Brain Mapping” (1364 papers), “Journal of Neural Engineering” (590 papers), “Frontiers in Brain Science Research” (341 papers), “Ieee Transactions on Neural Systems and Rehabilitation Engineering” (328 papers), and “Frontiers in Human Brain Science Research” (312 papers). These journals have published important articles on brain science research, particularly “Human Brain Mapping,” which not only has the highest publication volume but is also ranked in the JCR subject category Q1 zone and has a higher impact factor, making it the most important journal in the field of brain science research (Table 4). In terms of the number, frequency, and centrality of citations, high-impact factor journals such as “PNAS,” “NeuroImage,” and “Journal of Neural Engineering” are quite active in this field. Although they do not have many publications, they do have many collaborative citations. This means that the quality of their publications was far above the average (Table 5).

3.3.2 Literature Citations and Highly Cited Literature Analysis

References in an article demonstrate the continuity of scientific research and provide a scientific basis for the author's argument. We have summarized the references cited in articles published over the past decade. Table 6 shows the top 10 cited articles, of which nine are research papers, and one is a review article. The citation frequency of literature reflects the value of literature in the field of brain science research over the past decade, and highly cited papers constitute the knowledge foundation of brain science research. They not only provide a theoretical basis for further research at the forefront of brain science research technology but also lay the foundation for the application of brain science research scenarios. It is found that the paper “Deep Learning With Convolutional Neural Networks for EEG Decoding and Visualization” by Schirrmeister Robin Tibor has an important role in the field both as an article with high citations and as a highly cited article in the field (Tables 6 and 7). This indicates that Schirrmeister Robin Tibor and their Translate Neurotechnol Lab at the University of Freiburg may be the most promising and powerful innovation sources in the field of brain science research in the future.

3.5.1 “Brain Exploration,” the Exploration of the Function, Structure, and Principles of the Brain, Including the Physical Composition, Biological Mechanisms, and Working Functions of the Brain

Cluster #2 (diffusion tensor imaging), Cluster #4 (fMRI), and Cluster #7 (artifact removal) were applied to brain exploration and cognitive research (Figure 7). In terms of brain development, Cortina et al. (2021) used magnetic resonance relaxation and diffusion tensor imaging to explore gender and age-related microstructural differences in brain white matter. Bernard et al. (2022) discovered a synaptic signaling pathway for controlling the excitability of pyramidal cells and inhibitory interneurons expressing parvalbumin, providing the first evidence for protein synthesis regulation specificity in brain development. In terms of perception, researchers used optogenetic techniques to manipulate the ventral–posterior–medial thalamus of the tactile pathway in awake mice and measured the discharge activity of the thalamus and primary somatosensory cortex using extracellular electrophysiology and genetically encoded voltage imaging techniques (Borden et al. 2022).

In terms of learning and memory, researchers have inferred the possible existence of a retrograde mechanism that coordinates the aggregation of memory neurons in different brain regions (Lavi et al. 2023). Cai et al. (2021) confirmed that the difference in cortical thickness between mild cognitive impairment and degeneration is related to the upregulation of genes involved in chemical synaptic transmission. Tsai et al. (2022) proposed an adaptive online automatic speech recognition (ASR) technique that integrates Hebbian/anti-Hebbian neural networks into the ASR algorithm based on the adaptive pseudo shadow subspace reconstruction technique of Hebbian/anti-Hebbian learning networks to enhance the performance of brain–machine interfaces.

3.5.2 “Brain Protection,” Treatment of Brain-Related Diseases

Scientists are attempting to translate these discoveries into breakthroughs in treating human physiological and psychological ailments by gaining a deeper understanding of brain mechanisms and processes. Clusters #1 (stroke), #3 (amyotrophic lateral sclerosis), and #5 (awake surgery) (Figure 7) represent applications of brain science research in the field of disease treatment. Stroke and amyotrophic lateral sclerosis remain challenges in the urgent need for breakthroughs in brain science research. Bermudez I Badia et al. (2013) proposed in 2013 a hybrid BCI-virtual reality (VR) system that combines personalized motor training in a VR environment, exploiting brain mechanisms for action execution and observation, and a neurofeedback paradigm using mental imagery as a way to engage secondary or indirect pathways to access undamaged corticospinal tracts. Subsequently, BCI, MI-based BCI training, and MI and VR augmentation-based BCI training have been adopted to promote the recovery of upper limb function, cortical excitability, and cognitive task performance in stroke patients (Tao et al. 2023, Pichiorri et al. 2023, Kern et al. 2022, Aristela de Freitas et al. 2022). In 2014, Goljahani et al. (2014) proposed that preprocessing based on Bayesian single-trial event-related potential estimation technology provided feasibility for P300-based single-channel BCIs to improve operability and acceptability for patients with amyotrophic lateral sclerosis. In recent years, artificial intelligence technologies, such as interpretable convolutional neural networks and deep learning, have been used in BCIs to aid the movement of patients with amyotrophic lateral sclerosis. The brain–machine interface speller that encodes code-modulated visual evoked potentials (cVEPs) can enhance communication with patients with amyotrophic lateral sclerosis (Riccio et al. 2022, Aliakbaryhosseinabadi et al. 2021, Verbaarschot et al. 2021, Tajmirriahi et al. 2022). In 2013, Silva Paiva et al. (2013) described an approach guided by frameless neuronal activation and preoperative functional mapping with transcranial magnetic stimulation (TMS) for surgical planning. They verified complete control of seizures (Engel class 1A) in patients with refractory epilepsy. T. Li et al. (2015) used direct electrical stimulation for awake surgery for the localization and removal of gliomas responsible for language mapping. Direct electrical stimulation has been proven to be a reliable and noninvasive method for intraoperative language mapping, allowing for the safe removal of language areas in gliomas. Motomura et al. (2015) evaluated research showing that awake mapping can protect higher cognitive functions, including working memory and spatial cognition, in patients with nondominant right frontal lobe tumors. However, due to insufficient knowledge of the etiology and development of many brain diseases, many of them remain incurable.

3.5.3 “Brain Creation,” the Development of Brain Functions or the Borrowing of Brain Function Development Techniques

With a precise understanding of the brain's function and operational mechanisms, humans have begun to utilize and enhance the brain's capabilities in multiple ways, even by creating artificial brains. Clusters #0 (brain mapping) and #6 (functional connectivity) (Figure 7) were aimed at developing and applying brain functions. There are three main directions. First, in-depth research on brain–machine interface technology is conducted to build seamless, real-time, and side-effect-free brain–machine interfaces and enable perfect bidirectional communication between the brain and machines by timely recording and interpreting neural interaction patterns. In 2013, Rezakova et al. (2013) conducted research on dynamic mapping of the brain and cognitive control of virtual games using fMRI technology. Subsequently, new technologies such as extraction, emotion recognition, convolutional neural networks, transfer learning, brain modeling, and channel selection have been applied to the development of brain–machine interfaces (Luckett et al. 2023, Jami et al. 2022, Blood 2022, J. Zhang, Liu, et al. 2022).

In addition, achieving direct communication between the brain and machines has become a new path for enhancing human capabilities. Through technologies such as AR and VR, BCIs can assist in neural rehabilitation, create virtual sensations, and improve brain performance in learning, memory, attention, and other aspects. For example, Xu et al. (2023) proposed a framework for dividing sleep stages using multi-subsegment functional brain connections and utilized phase-locked values to construct brain networks to explore the mechanisms of brain functional connections. They will use graph convolutional networks for automatic sleep stage monitoring and develop an online BCI system, which has important potential for improving the application of sleep staging.

Third, research on neural mechanisms, cognitive structures, and brain–machine interfaces can contribute to breakthroughs in the field of artificial intelligence, leading to the design of new learning models that can mimic human brain neuron activity and ultimately lead to the development of more intelligent machines and robots. Huang et al. (2022) proposed an active learning framework called variational deep embedding-based active learning (VaDEAL) as a human-centric computing method to improve the accuracy of diagnosing pneumonia. S. G. Zhang, Chen, Zhang, et al. (2022) introduced augmented reality (AR) technology to present visual stimuli for SSVEP-BCI (BCI based on steady-state visual evoked potentials) and designed a robot grasping experiment to validate the applicability of the AR-BCI system. The results showed that the AR-SSVEP-NAO system was applicable to robot-grasping tasks. Technologies such as AR, novel imaging, neural monitoring, and regulation will promote rapid development in various subfields of brain exploration, protection, and creation. Neuromorphic computing and brain–machine intelligence will transition from “inspired by brain structure” to “balancing inspiration from brain structure and function,” develop “perception intelligence” and “cognitive intelligence” synergistically, and shift from “specialized intelligence” to “general intelligence.”

Keyword co-occurrence analysis is based on the detection of keywords with high frequency and rapid growth within a certain period of time. Keywords with high burst intensity are important indicators that reflect research hotspots, frontiers, and the latest trends. Twenty emergent words were identified (Figure 8). “Task analysis” had the highest burst intensity, reaching 46.92, and occurred between 2020 and 2023. “Deep learning” was the second highest, occurring between 2020 and 2023. Both are important research topics and applications in the field of machine learning. The early explosive growth of keyword citations mainly focused on brain disorders and cognitive control. With further research, research hotspots have become more diverse. Since 2018, brain-like artificial intelligence has emerged in the field of brain science research. Underlying theories and algorithm models in the field of artificial intelligence, such as “task analysis,” “deep learning,” “transfer learning,” “feature extraction,” and “neural network,” are gradually increasing, becoming the frontier and hot topic of current research in the field of brain science research. This explains the main source of China's recent progress in the field of brain science research: the rapid breakthrough in the field of scientific research driven by the extensive application of artificial intelligence and the use of advanced technologies such as artificial intelligence and big data algorithms to accelerate the filling of gaps in traditional brain science fields such as neural science, thus achieving an important achievement in the number of papers published, rising from sixth to second place globally in the past decade. The noteworthy points are the explosive references of keywords such as “task analysis” (2020–2023, 49.92), “deep learning” (2020–2023, 44.24), “brain modeling” (2020–2023, 43.17), “transfer learning” (2020–2023, 26.05), “emotion recognition” (2020–2023, 20.46), “convolutional neural network” (2021–2023, 31.65), “feature extraction” (2021–2023, 25.85), “neural network” (2021–2023, 20.58), and so on, which will continue to be the future research hotspots, as indicated by their sustained prevalence until 2023.