Wavelet Guided 3D Deep Model to improve Dental Microfracture Detection.

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Microfractures (cracks) are a significant cause of tooth loss in industrialized nations.
• Undetected tooth cracks can progress, leading to pain and eventual tooth loss.
• Previous methods using cone beam computed tomography (CBCT) for crack detection have shown limited success.

Purpose of the Study:

• To develop and evaluate a novel model for detecting cracked teeth using high-resolution (hr) CBCT scans.
• To improve the accuracy and reliability of diagnosing tooth fractures.
• To enhance early detection for better treatment planning and tooth preservation.

Main Methods:

• A deep convolutional neural network (CNN) based crack detection model was combined with signal enhancement techniques.
• Experiments were conducted on a dataset of 45 ex-vivo human teeth (31 cracked, 14 control).
• A 3D CNN model integrating wavelet-based features was trained to predict fracture probability maps, outperforming a 2D CNN approach.

Main Results:

• The proposed model demonstrated improved accuracy in detecting fractures on both micro-Computed Tomography and hr-CBCT scans.
• The CNN model successfully generated probability maps to identify fractured regions, differentiating cracked from control teeth.
• The developed solution outperformed oral and maxillofacial radiologists in fracture detection from hr-CBCT scans.

Conclusions:

• The combined signal enhancement and deep 3D CNN model significantly improves the detection of cracked teeth.
• This AI-driven approach offers superior performance compared to existing methods and human experts.
• Early and accurate diagnosis of tooth fractures can lead to improved treatments and increased tooth retention.