Automated dentition segmentation: 3D UNet-based approach with MIScnn framework.

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Digital workflows in dentistry necessitate segmentation of regions of interest from cone-beam computed tomography (CBCT) scans.
• Manual segmentation is time-consuming and costly, driving the need for automated solutions.
• Convolutional neural networks (CNNs) offer an efficient automated method for CBCT scan segmentation.

Purpose of the Study:

• To evaluate the efficacy of a 3D UNet-based CNN model for automated segmentation of maxillary and mandibular teeth from CBCT scans.
• To compare the performance of the automated method against traditional segmentation techniques.

Main Methods:

• A 3D UNet CNN model was implemented using the Medical Image Segmentation CNN framework.
• A dataset of 351 CBCT scans with manually segmented ground-truth labels was utilized.
• Data preprocessing, augmentation, and model training were performed to analyze CNN performance.

Main Results:

• The CNN model achieved high accuracy in segmenting maxillary and mandibular teeth.
• Average Dice Similarity Coefficient values were 91.83% for maxillary and 91.35% for mandibular teeth.
• Intersection over Union, precision, and recall metrics further validated the model's effectiveness.

Conclusions:

• The 3D UNet-based CNN model effectively automates the segmentation of dentition from CBCT scans.
• Automated segmentation using CNNs provides accurate and efficient results, surpassing manual methods.
• This technology holds significant potential for improving diagnostic and treatment planning processes in dentistry.