Artificial intelligence system for automatic tooth detection and numbering in the mixed dentition in CBCT

Authors:

ABSTRACT

Aim

To evaluate the effectiveness and accuracy of artificial intelligence (AI) by automating tooth segmentation in CBCT volumes of paediatric patients with mixed dentition, using nnU-Netv2 algorithm.

Background

Identifying and numbering teeth, the initial step in treatment planning, demands an efficient method.

Results

The accuracy, precision, and recall values for the successful numbering of deciduous and permanent teeth in CBCT scans were determined to be 0.99, 0.86, and 0.84, respectively. The values for the DC, Jaccard index, and 95% HD were calculated as 0.81, 0.81 and 1.93, respectively.

Conclusion

AI models offer a promising approach in the mixed dentition period and play a valuable role in dentists’ planning in terms of time and effort.

Study Design

In 49 CBCT scans, automatic segmentation and numbering of erupted/unerupted teeth of mixed dentition patients were performed using the CranioCatch labelling software (Eskisehir, Turkey). The dataset was randomly split into training (90%) and test (10%) groups. The developed model was trained with 1000 epochs using CBCT volumes and labelling. The performance of the model in numbering deciduous and permanent teeth was evaluated using several parameters, Dice Coefficient (DC), Jaccard index (Intersection over Union [IoU]).

PLUMX METRICS