Artificial Intelligence for Predicting the Aesthetic Component of the Index of Orthodontic Treatment Need

Abstract

The aesthetic component (AC) of the Index of Orthodontic Treatment Need (IOTN) is internationally recognized as a reliable and valid method for assessing aesthetic treatment need. The objective of this study is to use artificial intelligence (AI) to automate the AC assessment. A total of 1009 pre-treatment frontal intraoral photos with overjet values were collected. Each photo was graded by an experienced calibration clinician. The AI was trained using the intraoral images, overjet, and two other approaches. For Scheme 1, the training data were AC 1-10. For Scheme 2, the training data were either the two groups AC 1-5 and AC 6-10 or the three groups AC 1-4, AC 5-7, and AC 8-10. Sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were measured for all approaches. The performance was tested without overjet values as input. The intra-rater reliability for the grader, using kappa, was 0.84 (95% CI 0.76-0.93). Scheme 1 had 77% sensitivity, 88% specificity, 82% accuracy, 89% PPV, and 75% NPV in predicting the binary groups. All other schemes offered poor tradeoffs. Findings after omitting overjet and dataset supplementation results were mixed, depending upon perspective. We have developed deep learning-based algorithms that can predict treatment need based on IOTN-AC reference standards; this provides an adjunct to clinical assessment of dental aesthetics.

Keywords: Index of Orthodontic Treatment Need; aesthetic component; artificial intelligence.

Figure 1

The IOTN network with 2 inputs, an overjet module, a CNN module, and an output module.

Figure 2

Schematic of a training process. (a) Forward propagation: The model receives the first two inputs, namely the intraoral image and the overjet value. (b) Network generates a prediction: The network generates a prediction based on these inputs, aiming to learn the output of the IOTN. (c) Calculate loss function: The predicted IOTN value is compared to the gold standard (third input) to calculate the discrepancy. (d) Backward propagation and update network: The discrepancy is back propagated through each layer of the network to update their parameters.

Figure 3

Schematic of a test process. (a) Forward propagation: The model takes the two inputs, namely the intraoral image and the overjet value. (b) Network generates a prediction: The network predicts an IOTN-AC grade based on these inputs. (c) Performance Measurement: Performance is assessed by comparing the predicted value to the gold standard using diagnostic metrics, including SEN (sensitivity), SPE (specificity), PPV (positive predictive value), NPV (negative predictive value), and ACC (accuracy).

Figure 4

Summary of Scheme 0, Scheme 1, and Scheme 2: training and test. Need I or II is binary (IOTN 1–5 and 6–10). Need I, II, or III is ternary (IOTN 1–4, 5–7, and 8–10).

Figure 5

Scatter plots depicting binary prediction results for Scheme 1 (left) and Scheme 2 (right). In these plots, red region denotes IOTN-AC grades 1–5, while green region denotes IOTN-AC grades 6–10, as per the calibration clinician. Correct classifications are represented in the lower left and upper right quadrants of the plots.

Figure 6

Scatter plots for ternary prediction results of Scheme 1 (left) and Scheme 2 (right). In these plots, red region denotes IOTN-AC grades 1–4, yellow region denotes IOTN-AC grades 5–7, and green region denotes IOTN-AC grades 8–10, as per the calibration clinician. Outcomes in the lower left, center, and upper right regions are correct classifications.

Figure 7

Line graph representing the accuracy of all schemes.