Clinically Oriented CBCT Periapical Lesion Evaluation via 3D CNN Algorithm.

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Apical periodontitis (AP) is a common dental disorder, often asymptomatic and underdiagnosed.
• Accurate 3D lesion volume evaluation is crucial but manual segmentation of periapical lesions (PALs) on CBCT is time-consuming.
• Existing methods for PAL detection and segmentation lack efficiency and speed.

Purpose of the Study:

• To develop and validate a novel 3D deep convolutional neural network algorithm, PAL-Net, for rapid and accurate detection and segmentation of PALs.
• To assess PAL-Net's impact on diagnostic performance and time for dentists of varying experience levels.
• To evaluate the generalizability and robustness of PAL-Net across diverse datasets.

Main Methods:

• Development of a 3D deep convolutional neural network (PAL-Net) for automated PAL detection and segmentation on CBCT images.
• Internal validation using 5-fold cross-validation.
• External validation using datasets from Central, East, and North China.
• Evaluation of diagnostic performance using Area Under the Receiver Operating Characteristic Curve (AUC) and Dice Similarity Coefficient (DSC).

Main Results:

• PAL-Net achieved a high AUC of 0.98 in internal validation.
• The algorithm significantly improved dentists' diagnostic performance (AUC: 0.89-0.94 for junior, 0.91-0.93 for senior dentists) and reduced diagnostic time (69.3 min faster for junior, 32.4 min faster for senior dentists).
• PAL-Net demonstrated superior or comparable segmentation accuracy (average DSC > 0.87) to existing methods and strong robustness across external datasets.

Conclusions:

• PAL-Net offers a rapid, accurate, and robust solution for detecting and segmenting PALs on CBCT images.
• The algorithm enhances dental diagnostic efficiency and accuracy, providing valuable 3D volume information.
• PAL-Net has the potential to improve dental care, particularly in settings lacking expert radiologists or dentists.