Enhancing teeth segmentation using multifusion deep neural net in panoramic X-ray images.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Vision

Background:

• Dental panoramic X-ray image analysis is crucial but challenging due to poor contrast and image noise.
• Accurate teeth segmentation is vital for diagnosing dental conditions.
• Current methods are time-consuming and require manual intervention.

Purpose of the Study:

• To develop and evaluate a novel multi-fusion deep neural network for automatic teeth segmentation.
• To enhance the accuracy and efficiency of teeth region identification in panoramic X-rays.

Main Methods:

• Utilized a CNN-based encoder-decoder architecture with dual streams (conventional CNN and Atrous net).
• Implemented feature fusion at each encoding stage and dual-type skip connections.
• Employed deconvolutional layers in the decoder for precise segmentation map reconstruction.

Main Results:

• Achieved high accuracy (97.0% and 97.7%) on two independent datasets.
• Obtained excellent Intersection over Union (IoU) scores (91.1% and 90.2%) and Dice Coefficient Scores (DCS) (92.4% and 90.7%).
• Outperformed existing state-of-the-art deep learning models.

Conclusions:

• The proposed multi-fusion deep neural network offers a precise and automated solution for teeth segmentation.
• Demonstrates significant potential for improving dental disease diagnosis accuracy and efficiency in clinical practice.
• Requires fewer parameters than comparable models while achieving superior performance.