Artificial intelligence-based automated preprocessing and classification of impacted maxillary canines in panoramic radiographs

Abstract

Objectives: Automating the digital workflow for diagnosing impacted canines using panoramic radiographs (PRs) is challenging. This study explored feature extraction, automated cropping, and classification of impacted and nonimpacted canines as a first step.

Methods: A convolutional neural network with SqueezeNet architecture was first trained to classify two groups of PRs (91with and 91without impacted canines) on the MATLAB programming platform. Based on results, the need to crop the PRs was realized. Next, artificial intelligence (AI) detectors were trained to identify specific landmarks (maxillary central incisors, lateral incisors, canines, bicuspids, nasal area, and the mandibular ramus) on the PRs. Landmarks were then explored to guide cropping of the PRs. Finally, improvements in classification of automatically cropped PRs were studied.

Results: Without cropping, the area under the curve (AUC) of the receiver operating characteristic (ROC) curve for classifying impacted and nonimpacted canine was 84%. Landmark training showed that detectors could correctly identify upper central incisors and the ramus in ∼98% of PRs. The combined use of the mandibular ramus and maxillary central incisors as guides for cropping yielded the best results (∼10% incorrect cropping). When automatically cropped PRs were used, the AUC-ROC improved to 96%.

Conclusions: AI algorithms can be automated to preprocess PRs and improve the identification of impacted canines.

Keywords: artificial intelligence; automated algorithm; deep learning; impacted canine; panoramic radiographs.

Figure 1.

Schematic representation of the experimental workflow.

Figure 2.

Superimposed ROC curves of the validation results for the 1) as received PRs with (dashed red line) and without (dashed blue line) impacted canines; and the 2) cropped PRs with (solid red line) and without (dashed red line) impacted canines. ROC = receiver operating characteristic.

Figure 3.

Landmarks (maxillary central incisors, lateral incisors, canines, bicuspids, nasal area, and the mandibular ramus) labelled manually by expert clinician using MATLAB Image Labeller App (left), the same landmarks identified and labelled by trained AI detectors (right). AI = artificial intelligence.

Figure 4.

Example of ramal region (green box) and upper central incisors (red box) correctly identified by detectors and used for automatically selecting regions to crop out (region within the blue dashed line for the left side and region within the red dashed line border for right side).

Figure 5.

Conceptual representation of the key aspects and relations between the decision variables in the envisioned software (yellow box). Note that the work completed and reported in this manuscript is in green boxes, while work in progress is highlighted in pink boxes.