Trends of machine learning for dental caries research in Southeast Asia: insights from a bibliometric analysis

Introduction

Dental caries (DC), a common chronic oral disease, is a substantial threat to global health, with high incidence rates despite disease prevention efforts (Cheng et al., 2022). individuals encountering food insecurity are significantly more prone to dental caries, as evidenced by meta-analyses that demonstrate a higher risk of caries in food-insecure individuals compared to those with food security (Drumond, De Arruda, Bernabé, Mesquita, & Abreu, 2023). Both clinical researchers and dental practitioners are rigorously investigating optimal therapeutic interventions for this condition.

DC continue to be a major oral health concern in the Southeast Asia Region (SEAR), impacting children aged 5 to 15 (Kale, Kakodkar, Shetiya, & Rizwan, 2019). Studies have demonstrated an elevated prevalence of DC among children in SEAR countries, with variability among age groups and nations (Kale et al., 2019). In these investigations, the dentition status based on the 1997 WHO criteria was the most widely employed index for evaluating DC (Kale et al., 2019). Furthermore, the burden of DC extends to Cambodian preschool children, with a significant proportion developing carious lesions and pulpally affected teeth, underlining the critical need for intervention to address Early Childhood Caries (ECC) in the region (Turton, Chher, Sabbah, Durward, Hak, & Lailou, 2019). Efforts to integrate oral health into national health systems and establish uniform treatment standards are critical steps toward combating the prevalence of DC in the SEAR area (Acharya, Mathur, Tadakamadla, & Brand, 2024).

To increase diagnostic accuracy and prevent caries from being neglected, computer-based intelligent vision systems employing deep learning techniques have been developed, with a remarkable accuracy rate of 99.13% in automatic caries diagnosis based on periapical images (Imak, Celebi, Siddique, Turkoglu, Sengur, & Salam, 2022). In addition, innovations in caries detection include self-training-based approaches exploiting small sets of labeled images and enormous quantities of unlabeled images, boosting caries detection and segmentation performance without the requirement for extensive annotated data (Qayyum et al., 2023).

The current state of research in Southeast Asia regarding the application of machine learning techniques to DC research is lacking in comprehensive studies that incorporate multimodal data, including behavioral factors, radiographic imaging, and microbial analysis, to improve treatment outcomes and caries risk assessment (Jusman, Anam, Puspita, & Saleh, 2021; Velmurugadass, Rohith, Harish, & Reddy, 2023; Wu et al., 2022). While machine learning models such as Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) have been effectively utilized in dental caries classification research, there remains a dearth of understanding regarding the distinct challenges and determinants that impact DC within Southeast Asian communities, notably in nations such as Thailand and Indonesia (Ula, Anjani, Ulva, Sahputra, & Pratama, 2022; Wu et al., 2022). Researchers can better understand the complex etiology of DC in this region through the application machine learning techniques to a wide range of data sources, including genetic, behavioral, and environmental factors.

Bibliometric analysis is of minimal significance as it marginally enriches researchers’ understanding of the subject matter and vaguely suggests potential paths for further investigation. By conducting a thorough assessment of literature utilizing machine learning methodologies, we have been able to provide a comprehensive overview of the themes, patterns, and bibliometric attributes that presently define this body of literature during the period spanning from 2002 to 2024. The outcomes derived from our research unveil significant findings and propose innovative pathways for additional exploration within this rapidly expanding realm of knowledge acquisition and academic inquiry.

Methods

Results

Publication output

The final bibliometric investigation consisted of 246 papers after the screening process. The growth in the quantity of DC articles is visualized in Figure 1.

Figure 1. The distribution of articles by year.

Table 1 demonstrates the geographical distribution of the leading ten countries that have conducted research on DC. Malaysia emerged as the top contributor with 59 publications, succeeded by Indonesia with 37 publications and Thailand with 29 publications. The country with the highest ratio of Multiple Country Publications (MCP) was Malaysia at 0.407, followed by Singapore at 0.375 and Indonesia at 0.27. Supplementary Figure 1 (Extended data) showcases the network of countries that have collaborated on at least one research project.

Table 1. Top 10 countries that published in 2002–2024.

Rank	Country	Articles	SCP	MCP	Freq	MCP_Ratio
1	Malaysia	59	35	24	0.243	0.407
2	Indonesia	37	27	10	0.152	0.27
3	Thailand	29	24	5	0.119	0.172
4	China	17	0	17	0.07	1
5	Singapore	16	10	6	0.066	0.375
6	USA	7	0	7	0.029	1
7	Iran	6	0	6	0.025	1
8	United Kingdom	6	0	6	0.025	1
9	India	5	0	5	0.021	1
10	Japan	5	0	5	0.021	1

Supplementary Table 1 (Extended data) showcases the leading ten institutions that have shown high productivity in conducting research on DC. Among these institutions, Universiti Sains Malaysia stands out as the most prolific university with thirty-nine published articles, followed by Chiang Mai University with thirty-six, and the National University of Singapore with thirty publications.

The utilization of the VOSviewer tool (Leiden University) was employed for conducting co-authorship analysis, aiming to present a thorough depiction of the entire network of countries involved in DC research. The level of collaboration between two countries is determined by the quantity of publications they have co-authored. Out of the total of 8 authors, only 3 authors are interconnected, with each of them having published a minimum of 5 articles. The examination of authorship by country resulted in the identification of six distinct clusters, as illustrated in Supplementary Figure 2 (Extended data).

In the discourse concerning research on DC, a total of 1220 scholars collectively contributed to the publication of 246 scholarly articles. Within the realm of DC research, Table 2 provides a comprehensive overview of the top 10 most prolific authors in this field. Noteworthy among these authors is Li Y, who authored a total of eight scholarly papers on the subject matter.

Table 2. Top 10 authors that published in 2002–2024.

Rank	Authors	Articles
1	Li Y	8
2	Gupta R	6
3	Liu Y	6
4	Wantanajittikul K	6
5	Abu PAR	5
6	Chen CA	5
7	Chen SL	5
8	Chen TY	5
9	Fürst T	5
10	Ngoc VTN	5

Co-occurrence analysis of top 10 keywords

The 246 articles were divided into three keyword clusters. The top ten keywords utilized in the articles that were retrieved can be found in Table 3. It is worth mentioning that human, dental caries, article, child, and artificial intelligence were some of the predominant categories. The examination of the co-occurrence of the top 10 phrases revealed the research hotspots in the field of dental caries (Figure 2).

Table 3. Top 10 author keywords 2002–2024.

Rank	Keywords	Occurrences
1	Human	114
2	Humans	79
3	Dental caries	61
4	Article	54
5	Child	52
6	Female	48
7	Male	44
8	Artificial intelligence	41
9	Review	30
10	Adult	28

Figure 2. The co-occurrence network of the top 10 keywords in this research, 2002–2024.

Discussion

To the best of our knowledge, this research is the first bibliometric study to use machine learning to investigate DC. SEAR researchers are becoming more interested in using machine learning technologies to combat DC, according to this bibliometric analysis. The number of papers has increased over the previous ten years (>2014), which highlights the significance of this issue for public health as well as the potential of algorithms that use machine learning to provide assistance.

Machine learning has emerged as a powerful tool in combating dental caries, with Malaysia, Indonesia, and Thailand being at the forefront of utilizing this technology (Basri, Yazid, Zain, Yusof, Rani, & Zoolfakar, 2024; Wu et al., 2022; Wu et al., 2021). Research conducted in these countries has shown promising results in early caries detection and risk prediction through the application of machine learning algorithms such as artificial neural networks (ANN), convolutional neural networks (CNN), and statistical models (Basri et al., 2024; Sadegh-Zadeh, Rahmani Qeranqayeh, Benkhalifa, Dyke, Taylor, & Bagheri, 2022; Toledo Reyes et al., 2023; Wu et al., 2022). By integrating multimodal data, including genetic, microbial, demographic, and behavioral factors, these countries have been able to identify key predictors of caries development in both primary and permanent teeth, paving the way for more effective preventive measures and personalized treatment strategies (Sadegh-Zadeh et al., 2022; Toledo Reyes et al., 2023; Wu et al., 2022; Wu et al., 2021). The collaborative efforts between researchers and healthcare professionals in Malaysia, Indonesia, and Thailand highlight the potential of machine learning in revolutionizing oral health care and addressing the global burden of dental caries.

The research landscape on dental caries utilizing machine learning reveals a notable productivity ranking among institutions. Universiti Sains Malaysia emerges as the most prolific university in this field, boasting thirty-nine published articles, showcasing a strong commitment to advancing knowledge in dental caries detection. Following closely behind is Chiang Mai University with thirty-six publications, indicating a significant contribution to the research domain of dental caries and machine learning. Additionally, the National University of Singapore has also demonstrated substantial engagement in this area with thirty publications, further enriching the scholarly discourse on the early detection and management of dental caries through innovative technological approaches. These institutions’ dedication to research in dental caries using machine learning underscores the global effort to enhance diagnostic accuracy and treatment outcomes in oral healthcare.

In line with previous studies, the co-occurrence of the top 10 phrases in dental caries research, as highlighted in the bibliometric studies (Melo, Sanz, Forner, Rodríguez-Lozano, & Guerrero-Gironés, 2022; Patil et al., 2020), reveals a strong relationship with the predominant categories of human, dental caries, article, and child. Terms related to the validity of diagnostic methods, tools, and principles used in caries diagnosis, as well as general aspects of caries detection and diagnosis, are commonly found among the top phrases. This suggests that research in dental caries is heavily centered around human subjects, particularly children, with a significant emphasis on diagnostic methods and tools, as well as the overall understanding of the disease. However, interestingly, we found that the use of the keyword ‘artificial intelligence’ is on the rise.

Despite the fact that this bibliometric analysis provides interesting information, there are considerable limitations to take into account. The dependency on the Scopus database may result in the omission of significant articles from various sources. The correctness of the metadata associated with the articles also limits the study. Additionally, bibliometric analysis focuses on quantitative rather than qualitative evaluations of research findings and methodologies. Despite these limitations, the results demonstrate the increasing importance of employing machine learning to manage dental caries in Southeast Asia, which may guide future treatments and research areas.

Conclusion

We found the increasing evidence on the application of machine learning to mitigate the issue of dental caries in Southeast Asia. The bibliometric analysis identifies key authors, collaborative networks, and emerging topics, including dental caries and artificial intelligence. It also reveals growing research trends since 2014. This work underscores the significance of bibliometric analysis in tackling this critical public health challenge.