Detection of the separated endodontic instrument on periapical radiographs using a deep learning-based convolutional...

Area of Science:

• Dentistry
• Artificial Intelligence
• Medical Imaging

Background:

• Separated endodontic instruments are a common complication in root canal therapy.
• Accurate detection of these fragments is crucial for successful treatment and preventing further complications.
• Traditional radiographic interpretation can be challenging and time-consuming.

Purpose of the Study:

• To evaluate the diagnostic performance of an artificial intelligence (AI) system for detecting separated endodontic instruments.
• To assess the efficacy of the Mask R-CNN model in identifying and segmenting these fragments on periapical radiographs.

Main Methods:

• A dataset of 307 periapical radiographs was collected and divided into training (222 images) and testing (85 images) sets.
• The Mask R-CNN model was trained and tested using images labeled on the DentiAssist platform.
• The Intersection over Union (IoU) technique with an 80% threshold was employed for bounding box validation.

Main Results:

• The AI system achieved a mean average precision (mAP) of 98.809%.
• The model demonstrated high performance with a precision of 95.238%, recall of 98.765%, and an F1 score of 96.969%.
• The Mask R-CNN model successfully distinguished and segmented separated endodontic instruments.

Conclusions:

• The artificial intelligence system, specifically the Mask R-CNN model, exhibits excellent diagnostic performance in detecting separated endodontic instruments.
• AI-powered tools show significant potential to enhance the accuracy and efficiency of radiographic interpretation in endodontics.
• This technology can aid clinicians in identifying subtle or challenging cases of instrument separation.