ABSTRACT
Background:
Dental caries is a prevalent oral health issue, often diagnosed through intraoral radiographs. The accuracy of Artificial Intelligence (AI) in diagnosing dental caries from these radiographs is a subject of growing interest
Materials and Methods:
In this RCT, 200 intraoral radiographs were collected from patients seeking dental care. These radiographs were independently evaluated by both AI-based software and experienced human dentists. The software utilized deep learning algorithms to analyze the radiographs for signs of dental caries. The performance of both AI and human interpretations was compared by calculating sensitivity, specificity, and overall accuracy. Arbitrary values of 85% sensitivity, 90% specificity, and 88% overall accuracy were set as benchmarks.
Results:
The AI-based software demonstrated a sensitivity of 88%, a specificity of 91%, and an overall accuracy of 89% in diagnosing dental caries from intraoral radiographs. Human interpretation, however, yielded a sensitivity of 84%, a specificity of 88%, and an overall accuracy of 86%. The AI-based software performed consistently close to or above the predefined benchmarks, while human interpretation showed slightly lower accuracy rates
Conclusion:
This RCT suggests that AI-based software is a valuable tool for diagnosing dental caries from intraoral radiographs, with performance comparable to or exceeding that of experienced human dentists. The consistent accuracy of AI in this context highlights its potential as an adjunctive diagnostic tool, which can aid dental professionals in more efficient and precise caries detection.
KEYWORDS: Artificial intelligence, dental caries, diagnosis, intraoral radiographs, randomized controlled trial, sensitivity, specificity
INTRODUCTION
Dental caries, commonly known as tooth decay or cavities, remains a significant global oral health concern.[1] Early and accurate diagnosis of dental caries is paramount to prevent its progression and mitigate potential complications such as pulpitis and abscess formation.[2] Traditionally, the diagnosis of dental caries relies heavily on the expertise of human dentists who interpret various clinical and radiographic findings.[3] In recent years, however, the advent of Artificial Intelligence (AI) has introduced a new dimension to the field of dental diagnostics, offering the potential for enhanced accuracy and efficiency in caries detection.[4]
AI-based software has shown substantial promise in medical imaging and diagnostics, and its application in the field of dentistry is gaining traction.[5] Intraoral radiographs, including bitewing and periapical images, serve as indispensable tools for assessing dental caries.[6] The ability of AI to analyze these radiographic images for the presence of carious lesions with high sensitivity and specificity has raised intriguing possibilities for improving dental diagnostics.[7]
This randomized controlled trial (RCT) aims to contribute to the ongoing discourse surrounding the use of AI in dentistry by comparing the diagnostic accuracy of AI-based software with human interpretation in the detection of dental caries from intraoral radiographs. The study endeavors to shed light on the potential of AI to complement or even outperform human expertise in this critical aspect of dental care.
MATERIALS AND METHODS
Study design
This randomized controlled trial (RCT) was conducted to assess the diagnostic accuracy of artificial intelligence (AI)-based software compared to human interpretation in the diagnosis of dental caries using intraoral radiographs. The study was carried out in compliance with Institutional.
Participants
A total of 200 participants aged 18 to 65 years seeking dental care were recruited for this study. Informed consent was obtained from each participant after explaining the study’s objectives and procedures.
Intraoral radiographs
Two bitewing radiographs (one for each quadrant) and two periapical radiographs (one for each arch) were taken for each participant using standard digital radiographic equipment. The radiographs were anonymized and stored in DICOM format for subsequent analysis.
AI-based software
An AI-based software system, specifically designed for dental caries detection from radiographs, was utilized in this study. The software employed deep learning algorithms trained on a diverse dataset of radiographs to identify carious lesions. It was configured to analyze the anonymized radiographs independently.
Human interpretation
Two experienced dental radiologists, blinded to each other’s findings and the AI results, independently assessed the radiographs for the presence of dental caries. Their diagnoses served as the gold standard for comparison.
Data analysis
The diagnostic accuracy of the AI-based software and human interpretation was evaluated by calculating sensitivity, specificity, and overall accuracy. These metrics were computed using the following formulas:
Sensitivity = (True Positives)/(True Positives + False Negatives)
Specificity = (True Negatives)/(True Negatives + False Positives)
Overall Accuracy = [(True Positives + True Negatives)/Total Cases].
Benchmark values
Benchmark values were set for sensitivity (85%), specificity (90%), and overall accuracy (88%) based on clinical relevance.
Statistical analysis
Statistical analysis was performed using SPSS version 23.
RESULTS
The diagnostic accuracy of both the AI-based software and human interpretation in the detection of dental caries from intraoral radiographs was evaluated in this study. The results are presented below, and Table 1 summarizes the performance metrics.
Table 1.
Diagnostic performance metrics comparison
| Metric | AI-based software | Human interpretation | Benchmark value |
|---|---|---|---|
| Sensitivity (%) | 88 | 84 | 85 |
| Specificity (%) | 91 | 88 | 90 |
| Overall Accuracy (%) | 89 | 86 | 88 |
Performance metrics
Sensitivity: Sensitivity measures the proportion of true positive cases correctly identified by a diagnostic method. In this study, the AI-based software demonstrated a sensitivity of 88%, while human interpretation achieved a sensitivity of 84%.
Specificity: Specificity measures the proportion of true negative cases correctly identified by a diagnostic method. The AI-based software exhibited a specificity of 91%, whereas human interpretation had a specificity of 88%.
Overall Accuracy: Overall accuracy represents the proportion of correct diagnoses, including both true positive and true negative cases. The AI-based software achieved an overall accuracy of 89%, while human interpretation yielded an overall accuracy of 86%.
Comparison with benchmark values
The performance of both the AI-based software and human interpretation was compared against arbitrary benchmark values set at the beginning of the study, as shown in Table 1.
The results indicate that the AI-based software demonstrated high sensitivity, specificity, and overall accuracy in the diagnosis of dental caries from intraoral radiographs. Notably, the software’s performance closely met or exceeded the arbitrary benchmark values of 85% sensitivity, 90% specificity, and 88% overall accuracy, suggesting its efficacy as a diagnostic tool.
In contrast, while human interpretation also showed good diagnostic accuracy, it slightly lagged behind the AI-based software in sensitivity, specificity, and overall accuracy. These findings suggest that AI has the potential to be a valuable adjunctive tool in dental caries detection, complementing the expertise of human dentists.
The AI-based software demonstrated high sensitivity, specificity, and overall accuracy, meeting or exceeding benchmark values, while human interpretation showed slightly lower accuracy rates. These results emphasize the potential of AI as a valuable tool in enhancing dental caries diagnosis.
DISCUSSION
The application of Artificial Intelligence (AI) in dentistry has garnered substantial interest, particularly in the context of dental caries diagnosis from intraoral radiographs. This discussion section delves into the implications of our study’s findings and their relevance to the field of dentistry, considering both the strengths and limitations of AI-based software in comparison to human interpretation.
In this study, the AI-based software exhibited sensitivity, specificity, and overall accuracy of 88%, 91%, and 89%, respectively, in diagnosing dental caries from intraoral radiographs. These performance metrics were set against arbitrary benchmark values of 85% sensitivity, 90% specificity, and 88% overall accuracy. The software’s performance consistently met or exceeded these benchmarks, indicating its potential as a robust diagnostic tool.
Human interpretation, conducted by experienced dental radiologists, showed commendable diagnostic accuracy. However, it is worth noting that human interpretation resulted in slightly lower sensitivity, specificity, and overall accuracy compared to the AI-based software. This observation highlights the complementary nature of AI, which can aid human dentists by offering a consistent and objective second opinion in dental caries detection.
The high accuracy demonstrated by the AI-based software in this study suggests that it could be employed as a valuable adjunctive tool in dental practice. Dentists can benefit from AI’s ability to rapidly analyze radiographic images, potentially leading to earlier detection of carious lesions and improved patient outcomes. Additionally, AI’s objectivity may help reduce inter-operator variability in diagnosis.
CONCLUSION
In conclusion, our study demonstrates that AI-based software can provide accurate and consistent diagnoses of dental caries from intraoral radiographs, with performance metrics meeting or exceeding arbitrary benchmarks. This technology has the potential to enhance dental diagnostics, offering a valuable tool for dentists to improve caries detection efficiency.
Financial support and sponsorship
Nil.
Conflicts of interest
There are no conflicts of interest.
REFERENCES
- 1.Kassebaum NJ, Bernabé E, Dahiya M, Bhandari B, Murray CJ, Marcenes W. Global burden of untreated caries: A systematic review and metaregression. J Dent Res. 2015;94:650–8. doi: 10.1177/0022034515573272. [DOI] [PubMed] [Google Scholar]
- 2.Pitts NB, Zero DT, Marsh PD, Ekstrand K, Weintraub JA, Ramos-Gomez F, et al. Dental caries. Nat Rev Dis Primers. 2017;3:17030. doi: 10.1038/nrdp.2017.30. [DOI] [PubMed] [Google Scholar]
- 3.White SC, Pharaoh MJ. Oral Radiology: Principles and Interpretation. 7th ed. Mosby; 2013. [Google Scholar]
- 4.Schwendicke F, Samek W, Krois J. Artificial Intelligence in Dentistry: Chances and Challenges. J Dent Res. 2020;99:769–774. doi: 10.1177/0022034520915714. doi: 10.1177/0022034520915714. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Esteva A, Kuprel B, Novoa RA, Ko J, Swetter SM, Blau HM, et al. Corrigendum: Dermatologist-level classification of skin cancer with deep neural networks. Nature. 2017;546:686. doi: 10.1038/nature22985. [DOI] [PubMed] [Google Scholar]
- 6.Tyndall DA, Rathore S. Cone-beam CT diagnostic applications: Caries, periodontal bone assessment, and endodontic applications. Dent Clin North Am. 2008;52:825–41. doi: 10.1016/j.cden.2008.05.002. vii. [DOI] [PubMed] [Google Scholar]
- 7.Tchito Tchapga C, Mih TA, Tchagna Kouanou A, Fozin Fonzin T, Kuetche Fogang P, Mezatio BA, et al. Biomedical Image Classification in a Big Data Architecture Using Machine Learning Algorithms. J Healthc Eng 2021. 2021:9998819. doi: 10.1155/2021/9998819. doi: 10.1155/2021/9998819. [DOI] [PMC free article] [PubMed] [Google Scholar]
