Use of artificial intelligence in neurological disorders diagnosis: A scientometric study

Abstract

BACKGROUND

Artificial intelligence (AI) has become significantly integrated into healthcare, particularly in the diagnosing of neurological disorders. This advancement has enabled neurologists and physicians to diagnose conditions more quickly and effectively, ultimately benefiting patients.

AIM

To explore the current status and key highlights of AI-related articles in diagnosing of neurological disorders.

METHODS

A systematic literature review was conducted in the Web of Science Core Collection database using the following strategy: TS = ("Artificial Intelligence" OR "Computational Intelligence" OR "Machine Learning" OR "AI") AND TS = ("Neurological disorders" OR "CNS disorder" AND "diagnosis"). The search was limited to articles and reviews. Microsoft Excel 2019 and VOSviewer were utilized to identify major contributors, including authors, institutions, countries, and journals. Additionally, VOSviewer was employed to analyze and visualize current trends and hot topics through network visualization maps.

RESULTS

A total of 276 publications from 2000 to 2024 were retrieved. The United States, India, and China emerged as the top contributors in this field. Major institutions included Johns Hopkins University, King's College London, and Harvard Medical School. The most prolific author was U. Rajendra Acharya from the University of Southern Queensland (Australia). Among journals, IEEE Access, Scientific Reports, and Sensors were the most productive, while Frontiers in Neuroscience led in total citations. Central topics in AI-related articles on neurological disorders diagnosis included Alzheimer's disease, Parkinson's disease, dementia, epilepsy, autism, attention deficit hyperactivity disorder, and their intersections with deep learning and AI.

CONCLUSION

Research on AI's role in diagnosing neurological disorders is becoming widely recognized for its growing importance. AI shows promise in diagnosing various neurological disorders, yet requires further improvement and extensive future research.

Key Words: Artificial intelligence; Machine learning; Neurological disorders; Diagnosis; Bibliometric analysis

Core Tip: Artificial intelligence's (AI) role in diagnosing neurological disorders has been increasingly recognized. We conducted a scientometric analysis to explore the current status of articles in this field and identify the most prolific contributors from various perspectives. Johns Hopkins University in the United States emerged as the leading institution, with the United States also leading overall productivity in this field. IEEE Access was noted as the top journal. Research highlights AI's effectiveness in diagnosing diverse neurological disorders, offering significant benefits for patients and healthcare providers. Continued advancements are expected in AI's role in neurological disorder diagnosis.

INTRODUCTION

Artificial intelligence (AI) is the capability of computers to perform tasks typically requiring human intelligence[1]. AI and its various branches are being increasingly integrated into many aspects of business and society, including healthcare settings[2]. In healthcare, AI has generated a large amount of enthusiasm, especially through the development of precise machine-learning models[3]. Common applications of AI in healthcare include diagnosis, drug discovery, treatment, and enhancing doctor-patient communication[4]. Since AI is a relatively new technology, it is a vibrant area of research across all domains including medicine.

There is a growing interest in using AI to improve disease diagnosis[5]. Improvement in AI and computer vision show a good potential to significantly contribute to diagnostic specialties including radiology and pathology[6]. One of the most compelling applications of AI in healthcare is its ability to improve the accuracy of diagnosis and treatment. It can help healthcare providers in detecting symptoms earlier and more swiftly than many healthcare experts[7]. By using advanced branches of AI like machine learning and deep learning, AI can provide a more accurate understanding and predictions of disease behavior patterns; this would support clinical decision-making, enhance diagnostic accuracy, and reduce the physician’s workload[8].

AI has earned significant interest over the past decade, leading to increased research in the field. The AI application in diagnosis has proven to be an impressive novel factor in the diagnosis of critical illnesses and the prediction of disease prognosis[9]. For instance, AI can assist physicians in distinguishing between common and rare neurological disorders[10], such as Pompe disease, which is treatable if diagnosed early despite its bad progressive nature[11]. This development has been driven by an extensive collaboration between neuroscientists and AI experts, aiming to push the boundaries of neurological diagnosis to achieve early detection, a critical factor for successful treatment or disease progression prevention. Targeted treatments resulting from early diagnosis contribute to improving the quality of life for individuals affected by various neurological conditions.

Furthermore, AI is also capable of continuous analysis of diverse information coming promptly from various data sources like wearable devices, facilitating early diagnosis and treatment recommendations based on real-time information and current guidelines[12]. Most studies use machine learning for prognosis or diagnosis, although its application in treatment improvement remains limited[13]. Notably, some neurological machine learning algorithms have achieved impressive accuracy and precision rates, such as the VGG-19 model achieving 99.48% accuracy in magnetic resonance imaging (MRI) image classification, which offers a great classification of raw images without the need for manual extraction[14]. Another remarkable achievement includes the support vector machine (SVM), accurately predicting the progression of Alzheimer’s disease over a 4-year period with F1 scores of 88% for binary tasks and 72.8% for multitask scenarios[15]. These advancements underscore AI's potential to revolutionize neurological diagnosis and treatment, offering promising avenues for enhancing patient care and outcomes.

The main objective of this study is to conduct a comprehensive bibliometric analysis of AI applications in diagnosing neurological diseases, aiming to benefit both patients and healthcare providers. This review aims to serve as a foundational resource for anyone interested in the use of AI in neurological disease diagnosis, identifying key trends, top researchers, leading institutions, prominent countries, and frequently used keywords in this domain. By addressing these aspects, the study seeks to alleviate the challenges researchers face in navigating the literature on this topic. Ultimately, the insights gathered can advance the understanding, development, and dissemination of innovative diagnostic approaches, thereby improving patient outcomes. Furthermore, this study intends to highlight areas requiring further attention and research efforts within the field.

MATERIALS AND METHODS

Data collection

Data were retrieved from the Web of Science Core Collection (WoSCC) database on June 12, 2024, using the search criteria "full record and cited references" and "plain text". This database was selected due to its curation of high-quality, peer-reviewed literature from around the world[16]. The search strategy used the formula TS = ("Artificial Intelligence" OR "Computational Intelligence" OR "Machine Learning" OR "AI”) AND TS = ("Neurological disorders" OR "CNS disorders" AND "diagnosis"). Papers considered were limited to articles and reviews. To ensure the accurate inclusion of papers on the use of AI in diagnosing neurological disorders, all retrieved literature underwent screening based on their titles and abstracts. Any discrepancies were resolved through discussion until a consensus was reached.

The initial search yielded 471 articles, and after excluding entries outside the articles and reviews category, 381 articles remained. Subsequently, 276 articles were analyzed after excluding non-English articles and those unrelated to AI use in the diagnosis of neurological disorders. Detailed information about the screening process is illustrated in Figure 1.

Figure 1 Flowchart of data screening. AI: Artificial Intelligence; CNS: Central nervous system.

Statistical analysis

Microsoft Excel 2019 and VOSviewer (Centre for Science and Technology Studies, Leiden University, The Netherlands) were used to analyze all 276 publications. Full records for all publications were systematically extracted from WoSCC, including bibliometric parameters such as title, keywords, authors, countries, institutions, journals, citations, and publication year.

VOSviewer (1.6.20) was used to identify the primary contributors such as prolific authors, countries, and institutions. Furthermore, it was utilized to conduct keyword co-occurrence analysis, offering a comprehensive exploration of the scholarly landscape in the field. VOSviewer is a bibliometric software known for creating visualization maps, which display clusters and density colors[17]. Its algorithm ensures that frequently occurring terms are represented by larger bubbles, while terms with high similarity are positioned close to each other[18]. After extracting the data from the database, it was saved in .txt format and imported into VOSviewer. Next, we selected the type of analysis (co-authorship, co-occurrence, citation, or bibliographic coupling) and determined the unit of analysis, which varies depending on the chosen method (such as authors, countries, institutions, keywords, or sources). Finally, the data was visualized and further processed.

Microsoft Excel was used to organize the articles, prepare tables, and create a trend chart showing annual publications using its charting tools. It was also employed for data screening after extracting the data from the WoSCC database. Additionally, Excel helped present information about countries, institutions, authors, and journals in an organized manner. The data analysis results from VOSviewer were exported to Excel, where they were organized and structured to create the final tables.

RESULTS

Publications among years

Over the past few years, there has been an increase in the number of studies investigating the role of AI in diagnosing neurological disorders. This upward trend was particularly noticeable between 2021 and 2024 (Figure 2), reaching its peak in 2023 with 65 publications.

Figure 2 Number of publications among the years (2000-2024).

Distribution of authors

A total of 1611 authors contributed to articles exploring AI's role in diagnosing neurological disorders. The most prolific author, U Rajendra Acharya, has published 5 documents and is affiliated with the University of Southern Queensland, Australia, boasting the highest H-index among the top ten most productive authors. Following closely, the next eight authors each have three publications related to AI in neurological disorder diagnosis, representing diverse affiliations. Norlinah Mohamed Ibrahim and Khairiyah Mohamad are affiliated with the University Kebangsaan Malaysia, while the remaining authors hail from various institutions across Malaysia, Kuwait, the United Kingdom, and the United States, as detailed in Table 1. Figure 3A, generated using VOSviewer, illustrates a map where authors are depicted as nodes; larger nodes indicate higher publication counts. It also visualizes collaboration networks among authors studying AI's role in diagnosing neurological disorders. For instance, Mohammad Iqbal Omar is prominently connected to other authors such as Khairiyah Mohamad and Norlinah Mohamed Ibrahim, reflecting active collaboration in this field.

Figure 3 Visualization map of artificial intelligence articles in neurology disorders diagnosis. A: Visualization map of productive authors of artificial intelligence (AI) articles in neurology disorders diagnosis; B: Visualization map of productive institutions of AI articles in neurology disorders diagnosis; C: Visualization map of productive countries of AI articles in neurology disorders diagnosis; D: Network Visualization analysis of the keywords of AI articles in neurology disorders diagnosis.

Table 1 Top ten authors of artificial intelligence articles in neurology disorders diagnosis.

Rank	Authors	Documents	Country	H-index	Institute
1	U Rajendra Acharya	5	Australia	146	University of Southern Queensland
2	Norlinah Mohamed Ibrahim	3	Malaysia	39	University Kebangsaan Malaysia
3	Khairiyah Mohamad	3	Malaysia	-	University Kebangsaan Malaysia
4	M Murugappan	3	Kuwait	43	Kuwait College of Science and Technology
5	Mohammad Iqbal Omar	3	Malaysia	19	MERCY Malaysia
6	Ramaswamy Palaniappan	3	United Kingdom	39	University of Kent
7	Kenneth Sundaraj	3	Malaysia	30	Technical University of Malaysia, Malacca
8	Rjamanickam Yuvaraj	3	Malaysia	23	University Malaysia Perlis
9	Islem Rekik	3	United Kingdom	30	Imperial College London
10	Hojjat Adeli	2	United States	136	The Ohio State University

Distribution of institutions

Table 2 presents the top ten institutions contributing to AI articles on diagnosing neurological disorders. Johns Hopkins University (United States) and King's College London (United Kingdom) led with 7 articles each (2.54%), achieving the highest citations: 244 and 419 respectively. Following closely is Harvard Medical School (United States) with 6 articles (2.17%). Among these top ten institutions, four are in the United States: Johns Hopkins University, Harvard Medical School, University of Pennsylvania, and Boston Children's Hospital. Figure 3B illustrates the network of collaborations between institutions. Collaborations occur more frequently between geographically closer institutions, such as Harvard Medical School and Boston Children's Hospital, or between Johns Hopkins University and the University of California San Francisco.

Table 2 Top ten institutions of artificial intelligence articles in neurology disorders diagnosis.

Rank	Institute	Country	Documents	Percentage (n = 276)	Citations
1	Johns Hopkins University	United States	7	2.54%	244
2	King's College London	United Kingdom	7	2.54%	419
3	Harvard Medical School	United States	6	2.17%	34
4	Chinese Academy of Sciences	China	5	1.81%	131
5	University of Oxford	United Kingdom	5	1.81%	73
6	King Saudi University	Saudi Arabia	4	1.45%	40
7	University of Pennsylvania	United States	4	1.45%	132
8	University of Sao Paulo	Brazil	4	1.45%	37
9	Zhejiang University	China	4	1.45%	23
10	Boston Children's Hospital	United States	3	1.09%	4

Distribution of countries

Researchers from 62 countries participated in studies within this field, as detailed in Table 3. The United States ranked first as the most contributing country with 77 documents (27.9% of the total), followed by India with 42 documents (15.22%), and China with 36 documents (13%). Figure 3C displays the top 10 contributing countries. Many collaborations are evident between countries such as the United States, England, India, and Germany.

Table 3 Top ten countries of artificial intelligence articles in neurology disorders diagnosis.

Rank	Country	Documents	Percentages (n = 276)
1	United States	77	27.90%
2	India	42	15.22%
3	China	36	13.00%
4	United Kingdom	31	11.23%
5	Germany	22	7.97%
6	Italy	22	7.97%
7	Canada	15	5.43%
8	Australia	13	4.71%
9	Brazil	12	4.35%
10	Saudi Arabia	11	3.99%

Analysis of journals

Table 4 lists the top 10 active journals publishing articles on AI in the diagnosis of neurological disorders. The top 10 journals contributed 23.9% of the articles in the field. The three most prolific journals are IEEE Access with 12 articles, Scientific Reports with 10 articles, and Sensors with 8 articles. Regarding impact, Frontiers in Neuroscience ranks first with a total of 151 citations, followed by Plos One with 137 citations, and Sensors with 100 citations. Most journals are classified as Q1 or Q2, except for two journals-Frontiers in Computational Neuroscience and Frontiers in Neurology-which are classified as Q3 journals.

Table 4 Top ten journals of artificial intelligence articles in neurology disorders diagnosis.

Rank	Journal	Documents	Total citations	IF (JCR, 2024)	Quartile in category (JCR, 2024)
1	IEEE Access	12	78	3.4	Q2
2	Scientific Reports	10	96	3.8	Q1
3	Sensors	8	100	3.4	Q2
4	Applied Sciences-Basel	7	64	2.5	Q2
5	Frontiers in Neuroscience	7	151	3.2	Q2
6	Biomedical Signal Processing and Control	5	62	4.9	Q1
7	Plos One	5	137	2.9	Q1
8	Diagnostics	4	20	3.0	Q1
9	Frontiers in Computational Neuroscience	4	14	2.1	Q3
10	Frontiers in Neurology	4	13	2.7	Q3

IF: Impact factor; JCR: Journal Citation Reports.

Analysis of hotspots

The keyword map is derived from the frequency of keyword occurrences in the literature. Table 5 displays the top 20 keywords with high occurrence frequencies in AI-related articles on neurological disorders diagnosis. The most frequently occurring keyword is “machine learning” with 91 occurrences, followed by “classification” with 63 occurrences, and 'EEG' with 47 occurrences, highlighting the significant role of AI in diagnosing neurological disorders. Figure 3D illustrates the keyword occurrence network map, with each cluster representing a distinct research hotspot. In this study, six clusters were identified. Keywords such as “Alzheimer’s disease”, “Machine learning”, and “Artificial intelligence” formed one cluster (red). Another cluster (green) included “Parkinson's disease”, “Classification”, and “Speech”. Additionally, “EEG”, “Epilepsy”, and “Seizure detection” keywords were grouped in the blue cluster.

Table 5 Top 20 keywords of artificial intelligence articles in neurology disorders diagnosis.

Rank	Keyword	Occurrence
1	Machine learning	91
2	Classification	63
3	EEG	47
4	Alzheimer’s-disease	46
5	Deep learning	44
6	Parkinsons-disease	33
7	Diagnosis	32
8	Artificial intelligence	27
9	Epilepsy	26
10	Neurological disorders	23
11	Brain	16
12	Disease	16
13	Children	15
14	Features	15
15	Functional connectivity	15
16	MRI	15
17	Prediction	14
18	Dementia	12
19	Feature extraction	12
20	Performance	11

MRI: Magnetic resonance imaging; EEG: Electroencephalogram.

DISCUSSION

General

With the rapid evolution of AI and its use in healthcare, its application has significantly increased in recent years, particularly in diagnosing neurological disorders. Our study reveals that the number of published articles on this topic was relatively low between 2000 and 2019, but there has been a substantial increase in recent years. Articles published after 2019 constitute 81.2% of the total publications, indicating a surge in research during this period to explore advanced AI methods for supporting the diagnosis of neurological disorders[19]. AI has become increasingly crucial for neurologists as it enables the prediction of disease progression, facilitates adjustment of treatment plans, and ensures more accurate prognoses for patients[20].

Analysis on authors

The top three authors in this field were from Australia and Malaysia. Half of the top ten leading authors were from Malaysia, which can be attributed to the country's increased investment in education and research[21]. Malaysia's government and academic institutions have significantly bolstered research infrastructure and funding[22,23]. However, Malaysia was not among the top ten countries listed, and none of its institutions were among the top ten productive institutions. Regarding the H-index-a metric that assesses the quantity and quality of an author's publications-the most prolific author, U Rajendra Acharya (from Australia), had the highest H-index. Following him, Hojjat Adeli from the United States, despite being the least prolific among the top ten authors, had an H-index of 136, indicating his potentially higher impact and productivity compared to other authors.

Analysis of institutions & countries

Among the top ten productive institutions, four are located in the United States: Johns Hopkins University, Harvard Medical School, University of Pennsylvania, and Boston Children's Hospital. Together, these institutions contributed 7.2% of the total published literature. The United States has been identified as the most prolific country in articles on the role of AI in diagnosing neurological disorders. This observation is consistent with numerous bibliometric studies in healthcare and diagnostics, which consistently highlight the United States as a leading contributor in the field[7,9,24]. One of these studies also noted a significant presence of top institutions from the United States, corroborating our findings. This trend can be attributed to the United States's status as one of the wealthiest nations, with substantial investments and initiatives in advancing AI technologies[25]. Most of the top ten institutions and countries involved in such AI research in neurological disorders diagnosis are classified as high-income countries, which likely explains their extensive contributions, except for Brazil, classified as an upper-middle-income country, and India, classified as a lower-middle-income country according to the latest World Bank report (2024). The United States and the United Kingdom, which rank among the top five most productive countries in AI neurological research, have the most productive institutions in this field. This can largely be attributed to their substantial investments in research funding, their robust policy support for scientific innovation, and their commitment to fostering academic excellence. Both countries allocate significant budgets to support AI research, which enables them to attract top-tier researchers and facilitate high-impact collaborations. Moreover, their policies incentivize the growth of cutting-edge research in AI, creating a dynamic environment for innovation. Additionally, the ability of these countries to attract researchers from around the world for collaborative projects further enhances their research output. The research strength of the United States and United Kingdom in AI healthcare applications, in particular, is likely a result of these factors, which provide valuable insights for other countries seeking to develop their own research capacities and infrastructure[26].

Analysis on journals

IEEE Access and Scientific Reports journals were ranked as the top two most prolific journals listed as Q2 and Q1, respectively, by the Journal Citation Reports. This finding is consistent with another bibliometric study focused on AI in medicine, which identified IEEE Access as the most productive journal in the field[27]. This underscores the journal's significant impact and influence on AI within medical and neurological domains, as indicated by our study. Regarding impact, Frontiers in Neuroscience had the highest number of citations, highlighting its substantial contributions to the field. Researchers are encouraged to explore articles on AI in neurological diagnosis from these leading journals to establish a foundational knowledge base and stay abreast of the latest developments.

Research trends and frontiers

This study presented the key developments and roles of AI in diagnosing neurological disorders by analyzing a list of keywords that highlight current hotspots in the field.

The first cluster of keywords (red): Includes Alzheimer’s disease, artificial intelligence, machine learning, and stroke. AI, particularly machine learning, has significantly contributed to Alzheimer's imaging, aiding early diagnosis and guiding treatment efficacy assessments and strategies[28,29]. Such contribution includes: Utilizing fundamental machine learning architectures such as SVM, decision trees, and ensemble models[30]. Additionally, AI has been involved in stroke diagnosis, rehabilitation, and recommending optimal therapies[31].

The second cluster (green): Includes the following keywords: Parkinson's disease, computer-aided diagnosis, and dementia. Computer-aided diagnosis systems are recognized as AI tools that have witnessed growing use in recent years. They have made significant contributions, particularly in the detection and classification of Parkinson's disease[32,33]. One example includes a study that used biomedical sound measurements obtained from continuous phonation samples which were used as attributes in the diagnosis of Parkinson disease[34]. Various AI methods have also shown promise in early screening and detection of dementia, which can improve diagnosis and management of the disease's complications[35].

The third cluster (blue): Encompasses the following keywords: Electroencephalogram (EEG), epilepsy, seizure detection, and neural network. A neural network is a type of machine learning model. Several studies have employed neural networks to extract spatial characteristics from EEG data, leading to increased accuracy in epilepsy and seizure detection[36-38]. These studies consistently demonstrate improved epilepsy detection using this approach, which holds promise for enhancing the future management of seizures by neurologists. Another study utilized neural network systems, including 2D-CNN and LSTM, to detect and classify epilepsy seizures, and found high accuracy in both seizure detection and classification[39].

The fourth cluster (yellow): Encompasses the keywords: Autism, deep learning, and children. AI, particularly deep learning, has been utilized to assist in screening and diagnosing autism in children. Two studies examined the application of these AI models for early autism diagnosis using facial expressions of children[40,41]. One study employed five models to evaluate the accuracy of autism detection[40], while another utilized three models for diagnosis[41]. Both studies demonstrated more effective autism diagnosis, offering potential benefits to neurologists for early detection and management of the condition.

The fifth cluster (violet): Includes the keywords: Feature extraction, algorithm, and neurological diseases. Feature extraction is recognized as a method utilized in machine learning and AI tools. It has been increasingly employed alongside algorithms in neural network devices to aid in the diagnosis and treatment of brain and neurological diseases such as dementia, epilepsy, migraine, and autism[42].

The last cluster (light blue): Includes the keywords: Attention deficit hyperactivity disorder (ADHD), MRI, network, and depression. Adults with ADHD are three times more likely to develop depression than those without ADHD[43]. A study investigated the relationship between hippocampal function and levels of depressive symptoms in ADHD using MRI to assess resting-state functional connectivity of the hippocampus. The findings revealed that hippocampal abnormalities are linked to depressive symptoms, highlighting the need for further research in this area[44].

CONCLUSION

To our knowledge, this is the first scientometric paper focusing on the role of AI in diagnosing neurological disorders. This study offers valuable insights into research trends, potential collaborations, and cutting-edge topics in AI-related research on neurological disorders diagnosis. Over the past decade, interest in this field has significantly increased. According to our findings, U Rajendra Acharya, affiliated with an institution in Australia, emerged as the most productive author. Johns Hopkins University in the United States was identified as the most prolific institution. The United States also led in terms of overall productivity in this field. Among journals, IEEE Access stood out as the most productive journal. As AI technology continues to evolve, its applications are likely to expand to include more personalized diagnostic tools and predictive models for disease progression. The integration of AI with emerging technologies, such as personalized medicine and computer-aided diagnosis for movement disorders like Parkinson’s disease and dementia, will likely open new frontiers in neurological healthcare. Future research should focus on exploring these opportunities, as they have the potential to revolutionize how we diagnose, treat, and manage neurological diseases.