Deep Learning-Based Three-Dimensional Oral Conical Beam Computed Tomography for Diagnosis.

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Oral and maxillofacial surgical disease diagnosis relies on detailed imaging.
• Three-dimensional cone beam computed tomography (CBCT) provides crucial anatomical data.
• Limitations in image analysis can impact diagnostic accuracy.

Purpose of the Study:

• To develop and evaluate a deep learning-based algorithm for diagnosing oral and maxillofacial diseases using CBCT.
• To improve the accuracy of patient classification, lesion segmentation, and tooth size measurement from CBCT data.
• To reduce magnification errors in tooth measurements derived from CBCT.

Main Methods:

• A deep learning algorithm, DDOM (deep diagnosis of oral and maxillofacial diseases), was developed for patient classification, lesion segmentation, and tooth segmentation.
• The algorithm processed three-dimensional oral CBCT data.
• A correction equation involving the R value (distance from tooth center to FOV center) and vertical magnification rate was applied to reduce tooth size errors.

Main Results:

• The DDOM algorithm effectively performed patient-level classification and segmentation of oral and maxillofacial diseases.
• The proposed segmentation method accurately segmented individual teeth in CBCT images.
• Initial vertical magnification error averaged 7.4%, reduced to 1.0% after applying the correction equation.

Conclusions:

• Deep learning-based analysis of 3D oral CBCT data significantly enhances diagnostic capabilities.
• The DDOM algorithm assists clinicians in patient diagnosis, precise lesion localization, and effective surgical planning.
• Accurate tooth size measurements are achievable by correcting magnification errors in CBCT images.