The development of an image processing model to estimate tooth width and space requirements.

OBJECTIVES

The advent of machine and deep learning has raised the possibility of increased efficiency and indeed improved accuracy in relation to the interpretation of clinical images. Moreover, the use of deep learning has become imbedded within orthodontics with attempts to utilise this approach in informing treatment planning. The aim of this study was to develop and test an AI-based tool capable of measuring tooth width and dental crowding.

METHODOLOGY: Sets of pre and post treatment plaster study models were acquired of 245 patients who had treatment completed in a specialist orthodontic practice by a single orthodontist. These models were scanned to create stereolithography (STL) files. These files were used to train an AI tool capable of calculating the mesio-distal widths of each of the teeth in the arches. To test the validity of this model, direct comparison was made between the AI tool and human raters. A sample of twelve sets of pre-treatment study models were known as the "test data". The test data was analysed by a group of five orthodontic clinicians at two separate time intervals 6 weeks apart.

RESULTS

Statistical analysis confirmed a high level of agreement between human raters and also in relation to the novel AI-based tool. Inter-rater reliability among the human raters was high with Intraclass correlation coefficients (ICCs) ranging from 0.911 to 0.946 between the intervals. Manual measurements also demonstrated a high level of agreement with the AI tool (ICC= 0.956; Mean absolute difference= 0.4 mm).

CONCLUSIONS

The novel AI-based tool demonstrated reliability in calculating mesiodistal widths of teeth in an untreated arch. This tool will be further developed to incorporate other parameters such as arch levelling, to provide a means to calculate the overall space requirement within the dental arches.