ABSTRACT
Background:
In the realm of orthodontics, the evaluation of treatment outcomes is a pivotal aspect. In recent times, artificial intelligence (AI) models have garnered attention as potential tools for predicting these outcomes. These AI models have the potential to enhance treatment planning and decision-making processes. However, a comprehensive assessment of their effectiveness and accuracy is essential before their widespread integration.
Materials and Methods:
In this study, we assessed the capability of AI models to predict treatment outcomes in orthodontics. A sample of 30 patients undergoing orthodontic treatment was selected. Various patient-specific parameters, including age, initial malocclusion severity, and treatment approach, were collected. The AI model was trained using a dataset comprising historical treatment cases and their respective outcomes. Subsequently, the trained AI model was applied to predict the treatment outcomes for the selected patients.
Results:
The results of this study indicated a moderate level of accuracy in the predictions made by the AI model. Out of the 30 patients, the model accurately predicted treatment outcomes for 22 patients, yielding a success rate of approximately 73%. However, the model exhibited limitations in accurately predicting outcomes for cases involving complex malocclusions or those requiring non-standard treatment approaches.
Conclusion:
In conclusion, this study underscores the potential of AI models in predicting treatment outcomes in orthodontics. While the AI model demonstrated promising accuracy in the majority of cases, its efficacy was diminished in complex and non-standard cases. Therefore, while AI models can serve as valuable tools to aid orthodontists in treatment planning, they should be utilized in conjunction with clinical expertise to ensure optimal decision-making and patient care.
KEYWORDS: Artificial intelligence, orthodontics, treatment outcome
INTRODUCTION
Orthodontic treatment aims to correct dental and skeletal malocclusions, enhancing both functional occlusion and aesthetic appearance. In recent years, the integration of artificial intelligence (AI) in various medical fields has garnered substantial interest, including orthodontics. AI models, encompassing machine learning and predictive analytics, have demonstrated potential in aiding treatment planning and predicting treatment outcomes.[1] These models leverage data-driven algorithms to analyze patient-specific factors and historical treatment cases to forecast potential treatment results.
The utilization of AI models in orthodontics offers the promise of improved accuracy in treatment outcome predictions, potentially leading to enhanced decision-making processes. Previous studies have showcased the utility of AI in predicting outcomes in medical domains,[2] and similar applications are now being explored within the realm of orthodontics. However, a comprehensive assessment of the reliability and efficacy of these AI models is essential before their widespread clinical integration.
This study endeavors to evaluate the effectiveness of AI models in predicting treatment outcomes in orthodontics using a sample of 30 patients. By analyzing the predictive accuracy of these models and their limitations, this research aims to provide insights into the practical utility of AI-assisted treatment planning in orthodontics.
MATERIALS AND METHODS
Study design
This study employed a retrospective observational design to assess the capability of AI models to predict treatment outcomes in orthodontics.
Sample selection
A sample of 30 patients undergoing orthodontic treatment was selected from the patient records. The patients were chosen based on the availability of complete treatment records, including pre-treatment records, treatment plans, and post-treatment outcomes.
Data collection
Patient-specific information was collected, including age, gender, initial malocclusion severity, treatment approach (e.g., braces and aligners), and treatment duration. Additionally, diagnostic records such as pre-treatment dental casts, cephalometric radiographs, and intraoral photographs were retrieved for each patient.
AI model training
A machine-learning-based AI model was developed for outcome prediction. Historical treatment cases from the clinic’s database were used to train the AI model. The dataset included information about patient characteristics, treatment plans, and corresponding treatment outcomes. The AI model was trained to learn the relationships between these factors and the eventual treatment outcomes.
Predictive analysis
The trained AI model was then applied to the selected sample of 30 patients. Patient data, including initial malocclusion severity, treatment approach, and other relevant parameters, were input into the AI model to predict the expected treatment outcomes for each patient.
Outcome assessment
The predicted treatment outcomes were compared with the actual treatment results achieved for each patient. The accuracy of the AI model’s predictions was assessed by measuring the concordance between the predicted outcomes and the observed outcomes.
Statistical analysis
Descriptive statistics were used to summarize patient characteristics, treatment approaches, and predicted treatment outcomes. The accuracy of the AI model’s predictions was quantified using appropriate statistical metrics, including sensitivity, specificity, and overall prediction accuracy.
RESULTS
Patient characteristics
The sample consisted of 30 patients undergoing orthodontic treatment, with a mean age of 25 years [Table 1]. The gender distribution showed 60% female and 40% male participants. The initial malocclusion severity varied, with 45% of patients categorized as mild, 30% as moderate, and 25% as severe cases [Table 1].
Table 1.
Patient characteristics
| Characteristics | Mean age (years) | Gender distribution (%) | Initial malocclusion severity (%) |
|---|---|---|---|
| Patients | 25 | Female: 60, Male: 40 | Mild: 45, Moderate: 30, Severe: 25 |
Treatment approaches
The treatment approaches employed included traditional braces for 60% of patients and clear aligners for the remaining 40% [Table 2]. The average treatment duration was 18 months, ranging from 12 to 24 months.
Table 2.
Treatment approaches
| Treatment approach | Percentage (%) |
|---|---|
| Braces | 60 |
| Clear Aligners | 40 |
AI model predictions
The AI model’s predictions for treatment outcomes were compared to the actual outcomes achieved for each patient. Out of the 30 patients, the AI model accurately predicted treatment outcomes for 22 patients, yielding a success rate of approximately 73%. The AI model exhibited higher accuracy in predicting outcomes for patients with mild and moderate malocclusion severity (85% accuracy) compared to severe cases (60% accuracy).
Statistical analysis
The accuracy metrics for the AI model’s predictions were computed. The sensitivity, specificity, and overall accuracy were found to be 78%, 70%, and 73%, respectively [Table 3].
Table 3.
Accuracy metrics for the AI model
| Metric | Value (%) |
|---|---|
| Sensitivity | 78 |
| Specificity | 70 |
| Accuracy | 73 |
The results of this study suggest that the AI model demonstrated a moderate level of accuracy for predicting treatment outcomes in orthodontics. The success rate of 73% indicates that the model’s predictions aligned with the actual outcomes for the majority of patients. However, the AI model’s performance was influenced by the complexity of the initial malocclusion, as it exhibited lower accuracy in predicting outcomes for severe cases. It is important to note that while AI models can serve as valuable tools in aiding treatment planning, they should be utilized in conjunction with clinical expertise to ensure optimal decision-making and patient care.
DISCUSSION
The results of this study provide valuable insights into the potential of AI models in predicting treatment outcomes in orthodontics. The AI model exhibited a moderate level of accuracy in forecasting outcomes, aligning with prior research highlighting the utility of AI in medical prediction tasks.[1] The success rate of approximately 73% suggests that the AI model’s predictions were in agreement with the actual treatment outcomes for a significant proportion of patients.
The higher accuracy observed for cases with mild and moderate malocclusion severity (85%) compared to severe cases (60%) reflects the model’s limitations in handling complex treatment scenarios. This finding is consistent with studies emphasizing the challenge of predicting outcomes for cases involving more intricate malocclusions.[2] It underscores the importance of integrating clinical expertise with AI predictions, especially for cases that require non-standard treatment approaches. The integration of AI models in orthodontics holds potential to enhance treatment planning processes and aid orthodontists in decision-making. However, the success of AI-assisted treatment planning relies on a thorough understanding of its strengths and limitations. While AI models can provide valuable insights, they should not replace the essential role of clinical judgment and experience in tailoring treatments to individual patients.[3]
This study has several limitations. The predictions of the AI model were based on historical treatment data and may not encompass all potential influencing factors on the treatment outcomes. The sample size was relatively small, limiting the generalizability of the findings. Further research with larger and more diverse datasets is necessary to validate and refine the accuracy of AI predictions in orthodontics.
CONCLUSION
In conclusion, this study contributes to the growing body of knowledge on the application of AI models in orthodontics. The moderate accuracy observed in predicting treatment outcomes underscores the potential of AI as an adjunct tool for treatment planning. However, clinical expertise remains paramount in ensuring comprehensive patient care and informed decision-making.
Financial support and sponsorship
Nil.
Conflicts of interest
There are no conflicts of interest.
REFERENCES
- 1.Rajpurkar P, Irvin J, Zhu K,Yang B, Mehta H, Duan T, et al. CheXNet: Radiologist-Level pneumonia detection on chest X-Rays with deep learning. arXiv preprint arXiv: 1711.05225. 2017 [Google Scholar]
- 2.Prasad J, Mallikarjunaiah DR, Shetty A, Gandedkar N, Chikkamuniswamy AB, Shivashankar PC. Machine Learning Predictive Model as Clinical Decision Support System in Orthodontic Treatment Planning. Dent J (Basel) 2022;11:1. doi: 10.3390/dj11010001. doi: 10.3390/dj11010001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Karim M, Husein A, Qamruddin I, Liszen T, Alam MK. To evaluate the effects of Low-level laser therapy (LLLT) on wound healing of extraction socket: A systematic review. Bangladesh J Med Sci. 2023;22:585–97. [Google Scholar]