Deep-learning approach could help automate middle mesial canal detection

Researchers have developed a 3D convolutional neural network designed to identify middle mesial canals in mandibular molars on cone-beam computed tomography.

In a retrospective study published in the Journal of Endodontics, the researchers used 284 CBCT volumes and 28 cases to develop and test the 3D convolutional neural network — with endodontist-radiologist consensus establishing ground-truth middle mesial canal status. They also used a software to create multi-class masks of the dentin as well as the mesiobuccal, mesiolingual and middle mesial canals. Slice and case levels were evaluated for middle mesial canal detection.

The researchers noted excellent slice-level middle mesial canal detection and high agreement compared with individual observers and consensus references. Case-level middle mesial canal detection demonstrated favorable sensitivity, accuracy and 100% specificity. There were no statistically significant differences between either the 3D convolutional neural network and interobserver consensus reference at slice level or the 3D convolutional neural network and ground truth at the case level.

As a result of the findings, the researchers emphasized the potential accuracy of the deep-learning model in the automated detection of the middle mesial canal in mandibular molars.

Read more: Journal of Endodontics

The article presented here is intended to inform you about the broader media perspective on dentistry, regardless of its alignment with the ADA's stance. It is important to note that publication of an article does not imply the ADA's endorsement, agreement, or promotion of its content.