International publication trends in Biomedical Signal Processing research: A bibliometric analysis (2009-2018)

Abstract

Background This study aimed to analyze and assess the scientific outputs of biomedical signal processing by using bibliometric analysis.Methods Data were obtained from the WoSCC of Thomson Reuters, on January 21, 2019. VOSviewer (Leiden University, Van Eck and Waltman, Netherlands) and carrot 2 (Poznan University of Technology, Dawid Weiss, Poland) were used to analyze the knowledge maps and clusters of countries, research area and hot topics.Results A total of 335 articles on biomedical signal processing were identified. The number of publications increased only mildly during from 2009 (n=14) to 2018 (n=62). The majority of articles were published in the USA, and the leading institute was University of California System. Van Huffel S was the top authors on the topic, and the research area of “Engineering” generated the most publications. Cluter analysis (keywords and terms) indicated that “algorithm” and “extracted features” was the most hot topics on biomedical signal processing.Conclusion Overall, Through analysis of biosignal single processing related research in the past 10 years, the results found that the close international cooperation in this field, and the future research trends may be the signal acquisition methods and signal processing algorithms. They can provide reference for researchers in related field to choose research directions and find cooperative resources.

Background

Biomedical signal processing is an important research field in biomedical engineering, witch has significance research value in the fields of life science, clinical diagnosis, disease prevention and medical instrument production. Because of biomedical signals originate from human life activities, there are diversity and complexity in processing. As a result, this field has become the interdisciplinary category of many subjects and technologies, which includes clinical medicine, biology, pattern recognition, machine learning, signal modeling and simulation[1-6] .

Bibliometric analysis is a branch of library and information science that uses mathematical statistics to quantitatively describe, evaluate and predict academic status and development trends[7-8]. It has significant objective, quantitative and modeled macro research advantages[8]. At present, there are many professional softwares that can be used for bibliometric analysis[9]. In this study, VOSviewer was selected for visual analysis of bibliometrics, which can convert the large amount of literature data into a bibliometric map[10]. The researchers can understand the knowledge structure more directly, and find the hot topics and research areas from the maps. Biomedical signal processing have published a substantial amount of original research based on care digital signal processing, electrocardiosignal, and brain-computer interface, etc[11-13]. However, there are no bibliometric analyses that have explored research related to biomedical signal processing. This study used CitespaceIII to analyze biomedical signal processing articles retrieved on the Web of Science (Thomson Reuters Company) database and provides a retrospective and current view of the mainstream research on biomedical signal processing throughout the world.

Methods

All the data for this study were obtained from the Web of Science Core Collection (WoSCC) of Thomson Reuters, on January 21, 2019. The WoSCC, which includes the Social Sciences Citation Index, Current Chemical Reactions, and Index Chemicus, is the most frequently used source of scientific information. The search term “biomedical signal processing” was used to retrieve titles, keywords, author information, abstracts, and references published from 2009 to 2018. The following search string was used: TS=(“biomedical signal processing”) AND document type:(ARTICLE OR REVIEW). Data were downloaded from WoS in “Full record and cited references” and “Comma Separated Values (CSV)” formats.

The information statistics of this study mainly include authors, research institutions, countries and research fields, and the indicators were publication and citation. Microsoft Excel 2016 (Microsoft, Redmond, WA, USA) software tools were used to analyze the data[14]. VOSviewer 1.6.9 and carrot2 were used to analyze publication outputs and construct knowledge maps, to visualize the networks and clusters for countries, keywords and terms in biomedical signal processing[10, 15].

Results

Overall publication trends

A total of 335 studies on biomedical signal processing, published from 2009 to 2018, were retrieved from the WoSCC (Figure 1). Although the number of publications increased only mildly during from 2008 (n=14) to 2017 (n=62), the number of citation increased substantially from 2008 (n=5) to 2016 (n=1,081) (Figure 1). 269 studies (80.03%) were cited at least once, with an average of 13.08 citations per article for 4,381 total citations. We calculated the linear regression and found y=4.77x-7.27 with r²=0.884 (publication); y=125.66x-258.60 with r²=0.954(citation), respectively (Figure 2).

The most-cited articles

Among the 335 articles, the three most-cited articles were: “Surface EMG based muscle fatigue evaluation in biomechanics” by Mario Cifrek published in 2009 in Clinical biomechanics (196 citations, 17.82 citations per year)[16], “Zero-Velocity Detection-An Algorithm Evaluation” by Skog published in 2010 in IEEE Transactions on Biomedical Engineering (177 citations, 17.70 citations per year)[17], “Source Separation From Single-Channel Recordings by Combining Empirical-Mode Decomposition and Independent Component Analysis” by Mijović published in 2010 in IEEE Transactions on Biomedical Engineering (151 citations, 15.10 citations per year)[18]; See the Table 1 for the following 7 most-cited articles[19-25].

Countries or regions

These research studies were published by 49 countries or regions. The top 10 countries published 266 of the 335 studies, accounting for 79.40% of the total number of publications, which were cited 2,886 times. The country with the greatest number of publications was the USA (n=71), followed by Italy (n=31) and England (n=24); the country with the greatest number of citations was the USA (1,041 citations), Italy (341 citations) and Germany (272 citations); the country with the greatest per number of citations was USA (n=14.66 citations per article), Germany (12.95 citations per article) and Turkey (11.44 citations per article) (Table 2).

Countries or regions used in the 335 papers were analyzed by VOSviewer. The 57 countries or regions (defined as being used more than 5 times within documents in all of articles) were classified to 5 clusters (Figure 3):

1. Cluster 1--Austrilia (n=20), Belgium (n=15), Canada (n=20), Netherlands (n=10), People’s R China (n=21), Slovenia (n=5), South Korea (n=9), the USA (n=71);

2. Cluster 2--Austria (n=6), India (n=21), Romania (n=5), Singapore (n=7), Turkey (n=16);

3. Cluster 3--England (n=23), Finland (n=7), Germany (n=20), Poland (n=9), Taiwan (n=7);

4. Cluster 4--France (n=13), Italy (n=31), Switzerland (n=5);

5. Cluster 5--Brazil (n=7), Colombia (n=5), Spain (n=21).

Institutions and Authors

Among the 80 institutions, the top 10 institutions published 67 literatures, accounting for 20.00% of the total number of publications, which were cited 780 times. University of California System was the most publications (n=10), followed by the International Medical Equipment Collaborative (n=7) and KU Leuven (n=7), et al; In addition, the total number of authors retrieved in the 335 articles was 1,184, an average of 3.53 authors per article. The top ten authors accounted for 38 studies, 11.34% of the total, which were cited 685 times. Van Huffel S (n=6) was the most publications’ author, followed by Casciaro S (n=4) and Conversano F(n=4), et al (Table 3).

Research area analysis

These research studies were published by 52 research areas. The top 10 areas published 303 of the 335 studies, accounting for 90.45% of the total number of publications, which were cited 4,055 times. The top 3 number of publications were Engineering biomedical (n=149), Engineeeing electrical electronic (n=100) and Computer science interdisciplinary applications (n=46); the top 3 number of citations were Engineering biomedical (2,042 citations), Engineeeing electrical electronic (1,272 citations) and Computer science interdisciplinary applications (533 citations); The top 3 per number of citations were Mathematical computatonal biology (15.30 per article), Instruments instrumentation (14.63 per article) and Physiology (14.55 per article) (Table 4).

Hot hotspots

There were 335 papers on biomedical signal processing research had been included in this analysis. The visualization was generated by the Carrot2 system based on the 19 clusters. The top 5 clusters were Extracted features (n=53), Classification accuracy(n=37), Electrocardiogram ECG(n=36), Proposed model(n=34), Proposed design(n=31) (Figure 4).

In Figure 5, the left hand is the network of keywords, and the right hand is the network of keywords plus. In this study, keyword plus clustering results were used as the research direction of hot topics. Therefore, there are 1,114 keywords had been used to analyze by VOSviewer. The 53 keywords (defined as being used more than 5 times within documents in all of articles) were classified to 4 clusters:

I. Cluster 1(model and algorithm)--algorithm (n=31), frequency (n=8), model (n=11), neural-network (n=5), optimization (n=5), et al;

II. Cluster 2 (physiological signal recording)--brain (n=7), EEG (n=14), electrocadigram (n=9), recordings (n=12), signals (n=7), et al;

III. Cluster 3 (Signal transform and analysis)--decomposition (n=8), design (n=6), entropy (n=5), power (n=5), transform (n=15), et al;

IV. Cluster 4 (diagnosis of clinical diseases)--diagnosis (n=6), features (n=9), parkinsons-disease (n=5), reconstruction (n=13), validation (n=7), et al.

Discussion

A bibliometric analysis of biomedical signal processing research publications from WoSCC during the year 2009–2018 has been conducted. The analysis included three aspects: publication and citation trends, distribution of scientific research forces, research hotspots. Some findings were as follows:

Publication and citation trends

Biomedical signal processing research publication trends show that the publication of biomedical signal processing related articles increased slowly in the past decade, but its citation increased rapidly. This interest may be due to the number of basic research in this field is relatively small, but its application field is relatively wide, especially the interdisciplinary subject like biomedical signal processing.

From the 10 most-cited articles, their main research direction comes from the algorithm and mathematical model of biological signal processing. The most-cited is the research from Cifrek M, which summarized the signal methods and techniques from the standpiont of applicability to sEMG signals in fatigue-inducing situations relevant to the broad field biomechanics[16]; and this review as a basic konwleage had been mainly cited in research area of sports science, engineering biomedical and neurosciences, et al[26-28].

Distribution of scientific research forces

The USA published the most research on biomedical signal processing; this phenomenon is the same as many other research fields. This shows that the United States is a leader in the field. However, it is worth noting that German, Turkish and Italian research publications are small, but the average citation is high. In addition, although Peoples R China has the same publication as Germany, its quotation is the lowest in Top10, which indicates that the quality of academic papers in this field needs to be further improved.

From the perspective of national collaborative networks, the most connected clusters are between the United States and Canada.Other countries (regions) under the same cluster also include Austrilia, Belgium, Netherlands, People's R China, Slovenia, and South Korea, which are also the largest cluster in the collaboration network, indicating the frequent exchange of international scientific research cooperation among these countries (regions).

Research hotspots

The research hotspots mainly come from the clustering analysis of the subject words and key words included in the literature.

From the clustering analysis of subject words, the main research contents of these 5 clusters are as follows:

1) Extracted features: By extracting the features of human physiological signals, a feasible discriminant, diagnostic or classification model is established to accurately recognize the physiological state of the disease[29-31].

2) Classification accuracy: Through optimization algorithm, the accuracy of signal recognition model can be improved, including time series, wavelet analysis and machine learning[31-33].

3) Electrocardiogram ECG: Because of ECG is highly correlated with the diagnosis of many diseases, it has become an important part of the field of biomedical signal processing; research directions include signal monitoring, model algorithm optimization, disease screening, etc[34-36].

4) Proposed model: According to the data characteristics of biological signals, an algorithm model for signal recognition is constructed[37-38].

5) Proposed design: According to the requirement of biological signal acquisition, the hardwares and equipments were designed and improved[39-41].

From the clustering analysis of keywords, the main research contents of these 4 clusters are as follows:

1) Cluster 1 (model and algorithm): Establishment and optimization of mathematical models in the process of biological signal processing, including signal extraction, recognition and classification, and data conversion, etc[27, 29-30].

2) Cluster 2 (physiological signal recording): Signal acquisition and recording, including the design of signal acquisition hardware, the selection of recording methods and the determination of signal types, etc[39, 42-43].

3) Cluster 3 (Signal transform and analysis):Conversion of special signals or variables, such as signal processing of image data[32, 44-45].

4) Cluster 4 (diagnosis of clinical diseases): Based on the recognition and classification of signals, the diagnosis model of clinical diseases is established, which is mainly composed of ECG, EEG and image signals[46-48].

Conclusions

Overall, Through analysis of biomedical signal processing related research in the past 10 years, which can provide reference for researchers in related field to choose research directions and find cooperative resources. Major findings: (i) biomedical signal processing is an important link between biomedical engineering and clinical diagnostics; technological advances in this field may directly affect the diagnostic ability of future clinical diseases. (ii) International cooperation in this field is frequent, and scientific research is mainly distributed in North America and Europe. (iii) Signal acquisition methods and signal processing algorithms may be an important direction of future development, which deserves the attention of relevant professional researchers.

References

[1] Porta A , Nollo G , Faes L . Bridging the gap between the development of advanced biomedical signal processing tools and clinical practice. Physiological Measurement 2016; 36:627-631.

[2] Braojos R , Bortolotti D , Bartolini A , et al. A Synchronization-Based Hybrid-Memory Multi-Core Architecture for Energy-Efficient Biomedical Signal Processing. IEEE Transactions on Computers 2017; 66:575-585.

[3] Roy R N , Charbonnier S , Bonnet, Stéphane. Eye blink characterization from frontal EEG electrodes using source separation and pattern recognition algorithms. Biomedical Signal Processing and Control 2014; 14:256-264.

[4] Sun L , Chen B , Toh K A , et al. Sequential extreme learning machine incorporating survival error potential. Neurocomputing 2015; 155:194-204.

[5] Bruce EN. Biomedical Signal Processing and Signal Modeling. Expert Systems with Applications 2000; 38:6190–6201.

[6] Rubén Fraile, Kob M, Godino-Llorente JI, et al. Physical simulation of laryngeal disorders using a multiple-mass vocal fold model. Biomedical Signal Processing and Control 2012; 7:65-78.

[7] Li DC, Rastegar-Mojarad M, Okamoto J, et al. A Bibliometric Analysis on Cancer Population Science with Topic Modeling. Amia Jt Summits Transl Sci Proc 2015; 2015:102-106.

[8] Bornmann L , Mutz, Rüdiger. Growth rates of modern science: A bibliometric analysis based on the number of publications and cited references. Journal of the Association for Information Science and Technology 2015; 66:2215-2222.

[9] Xiaoxu Y, Junpeng Y, Jiping G, et al. The Comparative Study of Scientific Knowledge Mapping Tools Based on the Actual Case. Digital Library Forum 2014; 5: 66-71.(in Chinese)

[10] Eck NJV , Waltman L . Software survey: VOSviewer, a computer program for bibliometric mapping. Scientometrics 2010; 84:523-538.

[11] JuárezAguirre, Raúl, DomínguezNicolás, Saúl M, Manjarrez, Elías, et al. Digital Signal Processing by Virtual Instrumentation of a MEMS Magnetic Field Sensor for Biomedical Applications. Sensors 2013; 13:15068-15084.

[12] Polosin VG, Bodin ON, Ivanchukov AG, et al. Isoline Drift Correction in Digital Processing of the Electrocardiosignal. Biomedical Engineering 2016; 50:119-123.

[13] Liu G, Huang G, Meng J, et al. A frequency-weighted method combined with Common Spatial Patterns for electroencephalogram classification in brain-computer interface. Biomedical Signal Processing and Control 2010; 5:174-180.

[14] Lee YC, Chen C, Tsai XT. Visualizing the Knowledge Domain of Nanoparticle Drug Delivery Technologies: A Scientometric Review. Applied Sciences 2016; 6: 11.

[15] Stanisław Osiński, Dawid Weiss. Carrot²: Design of a Flexible and Efficient Web Information Retrieval Framework. International Atlantic Web Intelligence Conference 2003:240-249.

[16] Cifrek M , Medved V , Tonković S, et al. Surface EMG based muscle fatigue evaluation in biomechanics. Clinical biomechanics (Bristol, Avon) 2009; 24:0-340.

[17] Skog I , Handel P , Nilsson JO , et al. Zero-Velocity Detection—An Algorithm Evaluation. IEEE Transactions on Biomedical Engineering 2010; 57:2657-2666.

[18] Mijović Bogdan, De Vos M, Gligorijević Ivan, et al. Source separation from single-channel recordings by combining empirical-mode decomposition and independent component analysis. IEEE Transactions on Biomedical Engineering 2010; 57:2188-2196.

[19] Acar E , Dunlavy DM , Kolda TG, et al. Scalable tensor factorizations for incomplete data. 2011;106:41-56.

[20] Colominas MA, Schlotthauer Gastón, Torres María E. Improved complete ensemble EMD: A suitable tool for biomedical signal processing. Biomedical Signal Processing and Control 2014; 14:19-29.

[21] Ihlen EAF. Introduction to multifractal detrended fluctuation analysis in Matlab. Frontiers in Physiology 2012, 3:141.

[22] Min S , Lee B , Yoon S . Deep Learning in Bioinformatics. Briefings in Bioinformatics 2016; 18:851.

[23] Jin KH , Mccann MT, Froustey E, et al. Deep Convolutional Neural Network for Inverse Problems in Imaging. IEEE Transactions on Image Processing 2017; 26:4509-4522.

[24] Sazonov E S , Makeyev O , Schuckers S , et al. Automatic detection of swallowing events by acoustical means for applications of monitoring of ingestive behavior[J]. IEEE Transactions on Biomedical Engineering, 2010, 57(3):626-633.

[25] Chappell MA, Groves AR, Whitcher BJ, et al. Variational Bayesian Inference for a Nonlinear Forward Model. IEEE Transactions on Signal Processing 2009; 57:223-236.

[26] Evans NA , Emily D , Tim U . An electromyography study of muscular endurance during the posterior shoulder endurance test. Journal of Electromyography and Kinesiology 2018; 41:132-138.

[27] Karthick PA , Ghosh DM , Ramakrishnan S. Surface electromyography based muscle fatigue detection using high-resolution time-frequency methods and machine learning algorithms. Computer Methods & Programs in Biomedicine 2018; 154:45-56.

[28] Bigliassi M , Karageorghis CI , Nowicky AV , et al. Cerebral mechanisms underlying the effects of music during a fatiguing isometric ankle-dorsiflexion task. Psychophysiology 2016; 53: 1472-1483.

[29] Shahtalebi S , Mohammadi A . Bayesian Optimized Spectral Filters Coupled with Ternary ECOC for Single Trial EEG Classification. IEEE Transactions on Neural Systems and Rehabilitation Engineering 2018; 26: 2249-2259.

[30] Zille P , Calhoun VD , Wang YP . Enforcing Co-expression Within a Brain-Imaging Genomics Regression Framework. IEEE Transactions on Medical Imaging 2017; 37:2561-2571.

[31] Manish S , Bhurane AA , Rajendra AU . MMSFL-OWFB: A novel class of orthogonal wavelet filters for epileptic seizure detection. Knowledge-Based Systems 2018; 160: 265-277.

[32] Zafer C , Fatih KA , Velappan S. Prognostic model based on image-based time-frequency features and genetic algorithm for fetal hypoxia assessment. Computers in Biology and Medicine 2018; 99: 85-97.

[33] Ucar MK; Bozkurt MR; Bilgin C, et al. Automatic sleep staging in obstructive sleep apnea patients using photoplethysmography, heart rate variability signal and machine learning techniques. Neural Computing and Applications 2016; 29:1-16.

[34] Dai M, Xiao X, Chen X, et al. A low-power and miniaturized electrocardiograph data collection system with smart textile electrodes for monitoring of cardiac function[J]. Australasian Physical & Engineering Sciences in Medicine 2016; 39:1029-1040.

[35] Wedekind D , Kleyko D , Osipov E , et al. Robust Methods for Automated Selection of Cardiac Signals after Blind Source Separation. IEEE Transactions on Biomedical Engineering 2018; 65: 2248-2258.

[36] Mei Z, Gu X, Chen H, et al. Automatic Atrial Fibrillation Detection Based on Heart Rate Variability and Spectral Features. IEEE Access 2018; 6: 53566-53575.

[37] Manish S, Bhurane AA, Rajendra AU. MMSFL-OWFB: A novel class of orthogonal wavelet filters for epileptic seizure detection. Knowledge-Based Systems 2018; 160: 265-277.

[38] Roilhi F. Ibarra-Hernández, Miguel A. Alonso-Arévalo, Alejandro Cruz-Gutiérrez, et al. Design and evaluation of a parametric model for cardiac sounds. Computers in Biology and Medicine 2017; 89:170-180.

[39] Uktveris T; Jusas V. Development of a Modular Board for EEG Signal Acquisition. Sensors. 2018; 18: 2140.

[40] Algueta-Miguel JM , Blas CADLC , A. J. López-martín, et al. Design of CMOS amplifiers with offset rejection using positive-feedback QFG transistors. Analog Integrated Circuits and Signal Processing 2015; 85:217-221.

[41] Narasimhan S , Chiel HJ , Bhunia S . Ultra-low-power and robust digital-signal-processing hardware for implantable neural interface microsystems. IEEE Transactions on Biomedical Circuits & Systems 2011; 5:169-178.

[42] Kakareka JW , Faranesh AZ , Pursley RH , et al. Physiological Recording in the MRI Environment (PRiME): MRI-compatible hemodynamic recording system[J]. IEEE Journal of Translational Engineering in Health and Medicine 2018; 6: 4100112.

[43] Clifton DA, Wong D, Clifton L, et al. A Large-Scale Clinical Validation of an Integrated Monitoring System in the Emergency Department[J]. IEEE Journal of Biomedical and Health Informatics 2013;17:835-842.

[44] Abdelaal M, Elsarnagawy T. Analysis of Biomedical Signals with High Frequency Components of Wavelet Transform Compression Method on Electrocardiogram Signals[J]. Journal of Medical Imaging and Health Informatics 2018, 8: 785-788.

[45] Jin KH, Mccann MT, Froustey E, et al. Deep Convolutional Neural Network for Inverse Problems in Imaging. IEEE Transactions on Image Processing 2017; 21: 4721-4733.

[46] Valle BGD , Cash SS , Sodini CG . Low-Power, 8-Channel EEG Recorder and Seizure Detector ASIC for a Subdermal Implantable System. IEEE Transactions on Biomedical Circuits & Systems 2016, 10:1058-1067.

[47] Kumar SU , Inbarani HH . Neighborhood rough set based ECG signal classification for diagnosis of cardiac diseases. Soft Computing 2016; 21: 4721-4733.

[48] Cecília M. Costa, Silva IS , Sousa RDD , et al. The Association between Reconstructed Phase Space and Artificial Neural Networks for Vectorcardiographic Recognition of Myocardial Infarction. Journal of Electrocardiology 2018; 51: 443-449.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Availability of data and material

The datasets used and analyzed during the current study are available from the corresponding author upon reasonable requests.

Competing interests

The authors declare that they have no competing interests.

Funding

Not applicable.

Authors' contributions

LZ leaded the method application, experiment conduction and the result analysis. YC and PZ participated in the design of the research and the revision of the manuscript. KY provided theoretical guidance, the key term extraction program development and the revision of this paper. All authors read and approved the final manuscript.

Acknowledgements

Not applicable.

Tables

Table 1. Top 10 most-cited articles on biomedical signal processing

Table 2. Top 10 countries or regions on biomedical signal processing

Table 3. Top 10 authors and institution on biomedical signal processing

Table 4. Top 10 research areas on biomedical signal processing